Files
TREK/wiki/Plugin-Development.md
T
Maurice a87b656b43 3.3.0 (#1472)
* feat(plugins): grow the frame bridge — fill pages, confirm, openExternal, live context

- page/trip-page hosts pass fill: the frame pins to 100% height and ignores
  trek:resize, so a kit plugin's auto height report no longer collapses a full
  page into a floating island with dead space below (widgets keep self-sizing)
- trek:confirm renders the native ConfirmDialog host-side (the sandbox has no
  allow-modals) and answers trek:confirm:result; one at a time
- trek:openExternal opens validated http(s) URLs in a noopener tab — the
  sandbox has no allow-popups, so plugins simply couldn't link out before
- trek:notify accepts an optional duration, clamped to 1.5-15s
- context gains dir (rtl/ltr) and is re-pushed when locale or format settings
  change, not just on appearance mutations
- core events for the trip in view are forwarded as trek:event — names only,
  never payloads, mirroring the server-side events surface

* fix(plugins): move trip warnings out of the content area

The warning pills overlaid the top of every planner tab at full width, sitting
on the map and its toolbar. Now a warning from a plugin that owns a trip-page
tab renders as a compact chip in the navbar centre (click jumps to the tab; the
navbar centre is free on trip pages), and everything else floats above the
content at the bottom instead. Mobile keeps all warnings in the bottom overlay
since the desktop navbar isn't there. The trip tab's frame also opts into the
new fill mode.

* feat(plugin-sdk): 1.4.0 — motion library + new bridge helpers in the kit

Mirrors the host's animation vocabulary 1:1 into TREK_UI_CSS (menu/popover/
modal/backdrop/toast enters, drawer variant under 640px, page-enter, stagger,
skeleton shimmer, chart reveals) including the reduced-motion degrade to a
gentle fade. window.trek grows confirm(), openExternal(), onEvent() and a
notify duration, and applyContext now stamps lang/dir on the document so RTL
hosts get RTL plugin UIs.

* feat(plugins): surface registry download counts in browse

The registry now aggregates GitHub release download counts per plugin
(TREK-Plugins#18) as an entry-level downloadCount. Project it through
browse/detail and show it as a compact stat on the browse cards and in the
detail meta grid. Counts are raw asset downloads (updates and CI included),
so the UI says downloads, not installs.

* docs(plugins): document the grown bridge surface and motion classes

* fix(plugins): harden the new bridge paths

Review pass over the bridge additions:

- keep the unstable useToast() object out of the effect deps (ref instead) —
  it re-created the effect on every parent render, and with the new live
  repost that meant a trek:context flood into the frame
- reset loads/height/confirm state and key the iframe when a host swaps
  pluginId in place (tab bar, /plugins/:id) — the new plugin's document was
  refused as a 'navigated' frame and every kit promise hung
- confirm dialogs always lead with the host-controlled plugin name so a
  plugin can't dress its dialog up as a TREK system prompt; answer/refuse
  moved out of setState updaters (StrictMode ran them twice)
- Number.isFinite on the notify duration (NaN parked a sticky toast)
- don't forward other plugins' namespaced broadcasts as trek:event; a
  plugin's own plugin:{id}:* broadcasts now reach its frame though
- teach the SDK dev preview the confirm/openExternal contract so
  trek.confirm() resolves in /preview
- 999,950 downloads formats as 1M, not 1000k

* fix(planner): let plugin warning chips grow wider before truncating

The nav-centre chip capped at 340px, so a longer warning (e.g. the TREK x
Japan weather prompt) was ellipsised almost immediately. Scale it with the
viewport up to 520px so most messages read in full while still yielding on
narrow desktops.

* feat(plugins): sort the plugin browser by download count

Discover now honours the sort dropdown (it was always alphabetical) and adds
a 'Most downloads' option that ranks the registry by downloadCount. The sort
keys are scoped per tab — updates-first stays with Installed, most-downloads
with Discover — and snap back to name when the tab can't offer them.

* feat(plugin-sdk): auto-upgrade native <select> to a host-styled dropdown

A sandboxed plugin can't reach the host's components, and a native <select>
draws its popup from the OS — so plugin dropdowns never matched TREK. The design
kit now enhances every <select> into a keyboard-accessible listbox that uses the
kit tokens, keeping the real element as the value/form source (it still fires
change). Authors write a plain <select> and get the host look for free; opt a
field out with data-trek-native. validate warns when a plugin ships a <select>
without inlining the kit.

* feat(plugins): add issue url link

* feat(plugins): reservations write + cross-trip reads

- db:write:reservations -> reservations.create/update/delete, gated exactly like
  the REST/MCP path (reservation_edit + trip membership, acting user host-bound,
  no impersonation) and delegating to ReservationsService so the accommodation,
  budget-sync, booking-notification and reservation:* broadcasts match the web
  app 1:1 — a booking/flight/import plugin can finally write a reservation
- trips.listMine / reservations.listMine: enumerate every trip and booking the
  acting user can access (membership baked into listTrips, never a raw
  cross-tenant SELECT) — dashboards/aggregates were impossible before
- audit: derive auditability from METHOD_PERMISSION so a new capability method
  can't be added un-audited by omission
- typed ctx.reservations.* / ctx.trips.listMine, perm label (en/de), wiki

* feat(plugins): read scopes for journal, atlas, vacay and day notes

- db:read:journal / db:read:atlas / db:read:vacay expose the acting user's OWN
  journals / visited countries+regions / vacation plan across all their trips
  (user-scoped like costs.listMine, each gated on its addon being enabled),
  reusing the addon's existing readers
- db:read:daynotes -> daynotes.list(tripId, dayId), trip-scoped and
  membership-checked like the other trip reads
- typed ctx.journal / atlas / vacay / daynotes, perm labels (en/de), wiki,
  audit resource labels, tests

* feat(plugins): day notes write scope

- db:write:daynotes -> daynotes.create/update/delete, gated under the app's
  'day_edit' permission (like days) with the day verified to belong to the trip;
  reuses dayNoteService and broadcasts the same dayNote:* events so open
  sessions update live
- typed ctx.daynotes.create/update/delete, perm label (en/de), wiki, tests

* feat(plugins): run declared background jobs on a schedule

- plugins already declared jobs {id, schedule} but the cron was never wired. The
  host now schedules them: host-entry reports each job's schedule, the supervisor
  starts the jobs (node-cron) when the plugin goes active and stops them on
  kill/deactivate so nothing leaks
- opt-in via a new jobs:run permission — scheduled work runs with NO acting user
  (its trip reads stay refused; it can only use ctx.db and declared egress), so
  background execution is a distinct, admin-granted capability. Invalid crons are
  skipped and a throwing job can't break the host
- extracted a small, unit-tested scheduler (plugin-jobs.ts); perm label (en/de),
  wiki, tests

* feat(plugins): read scope for saved-place collections

- db:read:collections -> collections.listMine() / collections.get(id): the acting
  user's own collections (user-scoped, gated on the Collections addon), reusing
  collectionsService
- typed ctx.collections, perm label (en/de), wiki, audit resource labels, tests

* fix(plugins): translate new permission labels to all locales + cover the new wiring

- add the 8 new admin.plugins.perm.* labels (reservations/day-notes writes, the
  journal/atlas/vacay/day-notes/collections reads and jobs:run) to the remaining
  20 locales so the strict i18n key-parity test passes again
- cover the create-rpc-host reservation / day-note / cross-trip / addon-read deps
  (the real side-effect wiring the mocked rpc-host tests don't exercise) so the
  src/nest 80% branch-coverage gate holds

* feat(plugins): dev-link — hot-reload a local plugin against real data

Answers a plugin developer's ask: today you either get `trek-plugin-sdk dev`
(fast hot-reload but MOCK/fixture data) or the full build->pack->upload->activate
cycle (real data, no watcher). Neither gives "local dir + hot-reload + real data".

- POST /admin/plugins/link registers a plugin from a LOCAL built directory by
  symlinking it into the plugins volume — the loader already forks the resolved
  real path, so ZERO loader change — and registering it INACTIVE as `local:link`.
  Validates the manifest + refuses native binaries exactly like a sideload.
- POST /admin/plugins/:id/reload re-forks a linked plugin via the existing
  deactivate->activate primitive (same grants, no re-consent unless the manifest
  widened perms). A best-effort fs.watch auto-reloads on rebuild.
- It runs through the UNCHANGED capability RPC host: real, membership-gated data,
  acting user host-bound, no impersonation — code origin never touches the gate.
- Gated behind TREK_PLUGINS_DEV_LINK on top of admin + kill-switch, because a
  linked plugin bypasses the install-time signature model and, under `npm run
  dev`, the OS jail is off. Off by default; never reachable in production.
- discovery follows a symlinked <root>/<id>; uninstall/link never delete the
  author's source (link-safe removal for POSIX symlinks and Windows junctions).

* docs(plugins): document the dev-link real-data hot-reload workflow

Adds a "Test against a real instance's data (dev-link)" subsection next to the
mock-data SDK preview: TREK_PLUGINS_DEV_LINK, POST /link with a local built dir,
activate + consent, hot-reload via the file-watch / POST /:id/reload / Restart,
and the dev-only security caveats.

* feat(plugins): dev-link admin UI

- surface devLink (TREK_PLUGINS_DEV_LINK) in GET /admin/plugins so the panel shows
  the link form only where dev-link is enabled
- AdminPluginsPanel: a "Link a local plugin" form (path -> POST /link), a Dev-Link
  badge for source_repo=local:link, and adminApi.pluginLink/pluginReload
- fix: the plugin menu treated any non-local:upload source_repo as a GitHub repo,
  so a dev-linked plugin rendered github.com/local:link links — exclude local:link
- labels for the 6 new dev-link UI strings across all 22 locales

* docs(plugins): document the dev-link admin UI

The dev-link section showed only the curl call — surface the Admin → Plugins
"Link a local plugin" field (the primary path) and the Dev-link badge, with curl
kept as the scripting alternative.

* feat(plugins): enrich core events with { entity, entityId }

Subscribed plugins now learn WHICH entity changed, not just the event name — a
reservation/place/day/... id derived host-side from an explicit per-family
whitelist. Threaded through the six event hops WITHOUT touching actingUserId: the
handler still runs with no user, so the id is not dereferenceable (a trip read is
still refused; the id says what to react to, not what it contains). A non-entity id
can never surface — budget:member-paid-updated yields the itemId, never the userId
— and bulk/reorder/sub-entity payloads carry no id. The mapper is pure, synchronous
and never throws into the core broadcast. No new permission (reuses
events:subscribe); backend-only.

* feat(plugins): packing write scope with #858 privacy-scoped broadcasts

- db:write:packing -> packing.create/update/delete, gated under the app's
  'packing_edit' permission (like the REST path) with the host-bound acting user
  as owner; reuses packingService
- replicates the packing privacy model 1:1 (the controller/service helpers aren't
  exported): create/delete fan out to the item's viewers only (owner + recipients,
  or the whole room for a Common item); update runs the four public<->private
  transitions, dropping a freshly-privatized item from the room BEFORE re-adding it
  owner-only so it never leaks. A stale write is BAD_PARAMS with no broadcast
- typed ctx.packing.create/update/delete, perm label (22 locales), wiki, tests
  (rpc-host gating + the four transitions + owner-scoped delete)

* feat(plugins): tableContributor hook — host-rendered view columns/actions (backend)

The registry backend for plugin-contributed columns/actions in the native planner
views (the tabular-reservations use case), mirroring placeDetailProvider:
- hook:table-contributor + the tableContributor hook (getContributions(view,
  tripId, ctx) -> TableContribution[]), double-gated (implement + grant) like the
  other provider hooks
- GET /api/view-contributions/:view/:tripId — view whitelist + membership gate +
  per-provider timeout/fail-safe, plus the hardening the older provider hooks lack:
  every field is String-coerced + length-capped, kind/tone/target enum-whitelisted,
  per-provider counts capped (<=20 columns / <=10 actions), and a column url must be
  http/https/mailto (a javascript:/data: url is click-XSS into the native DOM)
- typed pluginsApi.viewContributions + the ViewContribution union, perm label
  (22 locales), wiki, hardening tests

* feat(plugins): render tableContributor columns/actions in the reservations view

The frontend for the tableContributor hook: a reusable PluginContributions layer
(usePluginViewContributions + PluginColumns/PluginActions) that renders the
host-normalized column/action leaves NATIVELY — a column is text/badge/link, an
action is a button that calls the plugin route or opens its sandboxed frame in a
modal (plugin markup only ever runs inside the opaque-origin iframe). Wired into the
reservations cards (both ReservationCard and TransitJourneyCard) as a strictly-
additive footer keyed by reservation id: zero change to a card when no plugin
contributes. Fetched once per view, fail-safe.

* docs(plugins): bring the permissions wikis current with this cycle

Plugin-Permissions.md was missing every permission added this cycle — add rows for
the read scopes (journal/atlas/vacay/daynotes/collections), the write scopes
(reservations/daynotes/packing, packing noting the #858 owner-scoping), jobs:run
and hook:table-contributor, and correct the events:subscribe row for the new
{ entity, entityId } hint. Add the jobs:run row to Plugin-Development.md too.

* fix(maps): stop quick one-finger pans zooming the map on mobile (#1440)

The global drag-drop-touch polyfill installs document-level touch listeners
on phones. On every single-finger touchend it records a timestamp, and if the
next touch starts within 500ms it synthesises a dblclick on the target, which
the map's default double-click-zoom turns into a zoom-in. Two quick one-finger
pans therefore zoomed instead of panning.

The polyfill only bridges HTML5 drag-and-drop to touch for planner reordering,
which is already disabled on mobile (#1432), so gate its import to viewports
>=1024px (the lg breakpoint useIsMobile uses). Removes the phantom-dblclick
source on phones while keeping touch DnD on large viewports; fixes both the
Leaflet and GL renderers.

* fix(feeds): emit TZID + VTIMEZONE so subscribed calendars respect time zones (#1453)

exportICS emitted timed DTSTART/DTEND as bare floating times (no Z, no TZID),
which iOS/Google Calendar render in the subscriber's local zone instead of the
zone TREK shows. Resolve an IANA zone per timed event — transport endpoints use
their stored timezone (departure drives DTSTART, arrival drives DTEND), while
assignments and hotel/restaurant reservations derive it from place coordinates
via tz-lookup — and attach TZID backed by a VTIMEZONE component. The all-trips
feed now carries deduped VTIMEZONE blocks so TZID references still resolve.

* feat(plugins): render tableContributor contributions in the places + day views

Extends the tableContributor frontend to all three planner views: hoist the shared
PluginCardFooter into PluginContributions, wire the places sidebar (keyed by place
id, rendered as a sibling after each row so the drag/scroll row stays untouched)
and the day panel (keyed by day id, guarded for a null day). Strictly additive +
fail-safe like the reservations view — nothing renders when no plugin contributes.

* fix(vacay): source holiday subdivisions from ISO 3166-2 so all states show

The state/region picker for public-holiday calendars was built from the
union of each holiday's counties for the current year, so a subdivision
only appeared if some holiday that year was tagged with it. States with
no state-specific holiday (e.g. US-WA, and AR/FL/NV/WY in 2026) silently
vanished, blocking calendar creation (#1456).

Source the full, correctly-named subdivision list per country from
ISO 3166-2 instead, merged with any nager county code ISO lacks. Only
region-partitioned countries get a picker, so nationwide-only countries
keep allowing a country-level calendar. No server change needed —
selecting a state already yields federal holidays via applyHolidayCalendars.

* fix(costs): settlements honor custom per-member splits (#1458)

calculateSettlement read each member's custom split amount but its query
never selected budget_item_members.amount, so hasCustomSplit was always
false and every settlement fell back to the equal split. Select bm.amount
so custom amounts drive the balances.

Also blank the Overview 'Per Person' / 'Per Person·Day' columns and CSV
for custom-split items, where a single averaged figure is meaningless.

* feat(plugins): read-convenience + todos + packing bags + tags + roster

A wave of small, high-value capabilities:
- weather:read (ctx.weather.get) — the host's cached forecast, tenant-free
- db:read:categories (ctx.categories.list) — the global place-category list
- db:read:tags / db:write:tags (ctx.tags) — the acting user's own tags, ownership
  re-checked before each write
- trips.members (ctx.trips.members) — the trip roster (id + display fields),
  membership-checked
- db:read:todos / db:write:todos (ctx.todos) — a trip's to-dos, gated by the app's
  packing_edit like the REST path, broadcasts todo:*
- packing bags on ctx.packing (listBags/createBag/updateBag/deleteBag/setBagMembers)
  under db:write:packing — no privacy, plain room broadcasts
perm labels (22 locales), both wikis, rpc-host gating + create-rpc-host wiring tests

* fix(dashboard): render next-trip boarding pass stats on Safari (#1459)

The boarding-pass bar carved its ticket-stub notches with a two-layer
radial-gradient mask composited via mask-composite: intersect (and legacy
-webkit-mask-composite: source-in). Safari mis-composites that multi-layer
path to fully transparent, hiding the entire stats bar while Chrome renders
it fine.

Split .hero-pass into an outer wrapper (left notch) and a .hero-pass-inner
glass panel (right notch), each carrying a single-layer mask so the
mask-composite path is never exercised. Renders identically across engines
and degrades safely where mask-image is unsupported.

* feat(plugins): write scopes for atlas, vacay, journal and collections

The write half of the user-scoped addon reads:
- db:write:atlas -> ctx.atlas.markCountry/unmarkCountry/markRegion/unmarkRegion +
  bucket-list create/delete. Every row is the acting user's own (visited_countries/
  visited_regions/bucket) — no trip scoping, no cross-tenant surface. Unblocks
  AirTrail-style two-way sync (#214)
- db:write:vacay -> ctx.vacay.toggleEntry/toggleCompanyHoliday. The plan is
  resolved HOST-SIDE from the acting user's active plan — a plugin can never name
  another plan, and toggleEntry only toggles the acting user's own PTO day
- db:write:journal -> ctx.journal.createEntry/updateEntry/deleteEntry, self-gated
  by journeyService.canEdit (owner/contributor) against the acting user
- db:write:collections -> ctx.collections.create/update/savePlace/copyToTrip/
  deletePlace, schema-validated; the service's per-collection role checks
  (assertAccess 404 / assertCanEdit 403) map onto RESOURCE_FORBIDDEN
All addon-gated, userless contexts refused, audited. Perm labels (22 locales),
both wikis, gating + wiring tests.

* fix(admin): name the Costs add-on consistently in the catalog

The budget add-on catalog entry still resolved to 'Budget' while the
feature is labeled 'Costs' everywhere else (trip tab, navbar). Align
admin.addons.catalog.budget.name with each locale's trip.tabs.budget
label. Closes #1464

* feat(plugins): file attach, collab content and gated member-add

- db:write:files -> ctx.files.create/createLink/update/softDelete under the app's
  separate file_upload/file_edit/file_delete rights. Content arrives as bounded
  base64 (10MB decoded cap, well under the app's 50MB), the extension is validated
  against the central blocklist BEFORE anything touches disk, and link targets
  must live on the same trip (findForeignLinkTarget). Broadcasts file:*
- db:write:collab -> ctx.collab.createNote/createPoll/votePoll/createMessage
  under collab_edit + the Collab addon, emitting the same collab:* events as the
  app; service-reported errors surface as BAD_PARAMS
- db:write:members -> ctx.trips.addMember. Adding a member GRANTS TRIP ACCESS, so
  it is deliberately its own permission behind the app's member_manage right
  (default: trip owner only) and never bundled with a lower-risk write; the acting
  user is recorded as the inviter, target must exist, owner/duplicate adds no-op
Perm labels (22 locales), both wikis, gating + wiring tests.

* fix(maps): honor check-in/out times for hotel bookend legs (#1465)

The day route drew the accommodation as the day's start/end whenever the
edge stop was a place, ignoring the morningIsSleptHere/eveningIsOvernight
provenance already computed by getDayBookendHotels. On a check-in day an
airport placed before check-in got a spurious hotel -> airport leg, and on
a check-out day a later "home" stop still got a home -> hotel return leg.

Add time-aware shouldDrawMorningLeg/shouldDrawEveningLeg helpers: the
morning leg is the home-base default on a check-in day but is dropped when
the first place is timed before check-in; the evening return leg is off on
a check-out day unless the last place is timed at/before check-out. Wire
them into the map polyline, the sidebar hotel connectors, and the Google
Maps export so all three stay consistent.

* feat(plugins): host-mediated notifications and LLM access

Two host-owned integration primitives — the plugin supplies intent, the host
owns the sensitive part:

- notify:send -> ctx.notify.send({title, body, link?, scope, targetId}). Delegates
  to notificationService.send with a new plugin_notification event (raw title/body
  carried as passthrough params), so recipient resolution, channel fan-out
  (bell inbox + email/ntfy/webhook) and per-user preferences all match core 1:1.
  Recipients are FORCED to the acting user (scope 'user', targetId === uid) or a
  trip they belong to (scope 'trip'); scope 'admin' refused; the in-app link must
  be a relative /path (open-redirect-safe). No arbitrary recipient, no impersonation.
  Users can mute plugin notifications like any other event.
- ai:invoke -> ctx.ai.complete(prompt) / ctx.ai.extract(text, jsonSchema). Runs the
  admin/user-configured provider via resolveLlmConfig + the existing extraction
  client under the acting user; the host holds the (encrypted) key, the plugin
  never sees it. Refused when no provider is configured; 20k-char caps. Output is
  DATA (complete -> {text}, extract -> {results}) and never auto-written, so
  prompt-injection can't reach a write without the plugin's own gated call.

plugin_notification wired through the shared NotificationEventKey + all 22 locales
(inbox passthrough + external channels). Perm labels (22 locales), both wikis,
gating + wiring tests.

* fix(budget): offer every Frankfurter-supported currency (#1470)

The cost currency picker was gated by a hardcoded 47-code list, so
currencies the app can actually convert (OMR, CRC, UGX, MKD, ALL, and
~115 more) couldn't be selected. Replace CURRENCIES/SYMBOLS with the full
set the Frankfurter v2 FX API supports (archived BGN/HRK dropped), unify
the dashboard offline fallback onto it, and teach currencyDecimals about
the newly reachable zero- and three-decimal currencies. A currenciesWith
helper keeps a previously saved (now-archived) selection selectable so it
isn't silently wiped.

* feat(plugins): tableContributor into the costs, packing and files views

Extends the shipped tableContributor hook to three more native views — no new
permission, no new attack surface: the same host-normalized, length-capped,
url-allowlisted (http/https/mailto), enum-bounded, fail-safe pipeline, just more
render sites.

- server: add costs/packing/files to the view-contributions whitelist
- client: widen the ViewName union + the api view type; render PluginCardFooter
  keyed by entityId in the budget category table (a colSpan footer row per item),
  the packing category group (footer after each item row, drag untouched) and the
  files list (footer after each row)

A currency plugin can now drop a converted-amount column onto a cost row, a
receipts plugin a 'view receipt' action onto a file, etc. Controller test asserts
the three new views are accepted; both wikis updated.

* fix(pdf): repeat day header on overflowing itinerary export pages (#1471)

* feat(plugins): map-marker provider hook — plugins can overlay trip-map markers

New declarative provider hook `mapMarkerProvider` (#587 "show bookings on map",
the single most-requested contribution class, with zero contribution point until
now):

- hook:map-marker-provider permission + MapMarkerProvider/MapMarkerContribution SDK
  types + HOOK_PERMISSION wiring
- GET /api/map-markers/:tripId (MapMarkersController) mirrors the view-contributions
  hardening: membership-gated, providers invoked host->plugin on a 5s timeout,
  fail-safe. Every field normalized server-side — coordinates range-checked
  (-90..90 / -180..180), strings String-coerced + length-capped, icon/tone enum-
  whitelisted, popup url http/https/mailto only (a javascript:/data: url would be
  click-XSS), marker count capped at 200 per plugin
- client: PluginMapMarkers layer renders the markers as plain Leaflet Marker+Popup
  inside the trip map; plugin JS NEVER runs on the map canvas, every value is
  host-vetted data. Threaded tripId through MapView; fail-safe fetch

Declarative-only by design, mirroring placeDetailProvider/tableContributor. Perm
label (22 locales), controller hardening test, both wikis.

* feat(plugins): show page plugins in the mobile bottom nav

Page plugins were reachable from the desktop nav pill (Navbar) but not the mobile
tab bar — you had to type /plugins/:id. BottomNav now reads page plugins from the
plugin store and appends them the same way global addons are, mirroring Navbar.
One-file client nav wiring; no new capability surface.

* feat(plugins): per-user plugin settings form + ctx.settings runtime read

Users can now enter their own per-plugin config (an API key, a preference) —
the prerequisite for almost every real integration, previously unreachable
(scope:'user' settings were only listed read-only in the admin panel).

- migration: plugin_user_config (plugin_id, user_id, config JSON) — each user's
  own values, separate from the admin-owned instance plugins.config
- PluginsService.getUserConfig / updateUserConfig / getUserConfigDecrypted +
  readUserSettingDecrypted: secrets encrypted at rest (apiKeyCrypto), masked to
  the client, an unchanged secret (the mask) keeps its stored ciphertext, and only
  DECLARED scope:'user' keys are ever stored
- GET/POST /api/plugin-settings/:id (PluginUserSettingsController) — its own
  user-gated path (not the admin surface, not the /:id/* proxy), JwtAuthGuard only,
  scoped to the acting user
- runtime: ctx.settings.get(key) -> the acting user's decrypted value (unconditional
  RPC, not sensitive cross-tenant; userless job/onLoad gets undefined)
- client: a Plugins tab in Settings host-renders each active plugin's scope:'user'
  fields as an editable form (secrets write-only), reusing the declarative field
  shape — no plugin markup executes

i18n (22 locales), wiki, rpc-host + service + masking/encryption tests.

* fix(journey): keep skeleton suggestions in sync with linked trip places (#1473)

Journey skeleton suggestions mirror a linked trip's day-assigned places, but
sync relied on scattered per-event hooks that several assignment mutation paths
never called: unassign, move and time-change fired nothing, no remove-on-unassign
capability existed, and every MCP assignment tool synced nothing. Skeletons drifted
from the trip.

Add an idempotent reconcileTripSkeletons(tripId) that re-mirrors the trip's
day-assigned places onto every linked journey (add missing skeletons, refresh
date/time/location on move, remove skeletons for unassigned places; filled entries
are detached + noted, never destroyed). Call it from every REST assignment handler
and MCP assignment tool, and fire onPlaceDeleted on single MCP delete_place for
parity. Extract a shared insertSkeletonEntry helper.

* fix(memories): drop hidden Immich assets so Live Photo motion parts don't show a broken thumbnail (#1474)

* fix(transit): anchor arrive-by search time to the destination timezone (#1479)

* feat(plugins): host-brokered OAuth client + trustworthy inbound webhooks

Two integration primitives where the host owns the sensitive part.

Trustworthy webhooks:
- auth:false routes now receive req.headers, but ONLY an explicit, credential-free
  allowlist (the common provider signature/event headers — stripe-signature,
  x-hub-signature-256, svix-*, x-gitlab-event, …). Cookie/Authorization/X-Socket-Id
  and every session/forwarded-auth header are stripped; authenticated routes get {}.
  A plugin can finally verify a provider signature without any way to leak a session.

Host-brokered outbound OAuth (oauth:client):
- the HOST runs the whole flow — authorize -> callback -> token exchange -> refresh —
  with PKCE + single-use, user-bound, TTL'd state, and HOLDS the tokens. The client
  secret + refresh token never leave the host; the plugin only triggers connect and
  reads a short-lived access token via ctx.oauth.getAccessToken() for the acting user.
- provider config (authorize/token url + scopes + client id/secret) is the plugin's
  admin-owned instance settings; endpoints must be https (SSRF backstop, private/local
  hosts refused). Tokens per-user + encrypted at rest (apiKeyCrypto).
- GET/POST /api/plugin-oauth/:id/{status,connect,callback,disconnect} — JwtAuth-gated,
  the callback always redirects to an in-app /settings path (never leaks an error).
- Settings -> Plugins gains a Connect/Disconnect control per configured plugin.

migration: plugin_oauth_tokens + plugin_oauth_state. Perm labels + form strings
(22 locales), both wikis, service (PKCE/state/exchange/refresh/encrypt) + controller
+ proxy header-allowlist + rpc-host gating + create-rpc-host wiring tests.

* fix(navbar): re-measure sliding tab pill after font load and resize (#1481)

The active tab pill was measured once in a layout effect keyed only on activeTab, so on a hard reload it captured the active (bold) label's width against fallback-font metrics and never re-ran when the web font swapped in, leaving the pill slightly offset.

Re-measure after document.fonts.ready resolves and on ResizeObserver changes (container + active button), with an idempotent state update to avoid redundant renders.

* fix(collections): keep the Add-place button reachable after the first save

On a wide/desktop layout the collection toolbar (which hosts the Add
button) was gated on !mapOverlay, so it unmounted as soon as the list
gained its first place with coordinates — leaving only an easy-to-miss
"+" in the map overlay. Keep the toolbar rendered whenever the user can
add a place, and drop the now-redundant map-overlay Add button so there
is a single, predictable Add affordance in every state.

Fixes #1485

* feat(plugins): days + accommodations reads/writes, endpoints on the reservation write path

Community feedback on the 3.2.1 plugin surface: a plugin could write days but
never list them (no way to learn day ids), day_accommodations had no surface at
all, and trips.getReservations was the one reservation read that dropped the
endpoints/day_positions hydration.

- trips.getDays / trips.getAccommodations under db:read:trips (tripRead gate),
  wired to the same dayService lists the REST GETs use
- trips.getReservations now returns the hydrated REST-parity list (endpoints,
  day_positions, joins, normalized accommodation_id) - strict superset
- new db:write:accommodations scope: ctx.accommodations create/update/delete
  gated by day_edit like the accommodations REST path, with the partner-hotel
  reservation + delete cascade and broadcasts intact
- reservation create/update pin the endpoints shape up front (BadParams instead
  of a mid-transaction NOT-NULL or a silently dropped row)
- perm label in all 22 locales, consent PERM_KEYS, wiki tables

* feat(plugins): day-detail widget slot in the day panel

Widgets can now mount inside the trip planner's day panel
(capabilities.widget.slot: 'day-detail'), scoped to the open day via a dayId in
trek:context - the same pattern as the place-detail slot. Covers the requested
per-day plugin content (logistics, outfit planning, live flight status) without
a new plugin type. Day-detail widgets stay off the dashboard, the consent panel
labels the slot in all 22 locales.

* feat(plugins): let the frame CSP serve a plugin's own static assets

The sandboxed frame runs at an opaque origin, so script-src 'self' never
matched and a plugin's own <script src>/<link> files were blocked - authors had
to inline entire React builds into index.html. Add a scheme-less host-source
pinned to the plugin's own /plugin-frame/<id>/ path (charset-checked Host +
plugin id so a stray token can't widen the policy; malformed Host falls back to
inline-only). Multi-file client builds now load as-is; remote hosts stay
blocked, so script URLs remain useless as an egress channel.

* fix(plugin-sdk): catch the package up to the server capability surface

The npm SDK's validator still knew only the 3.2.1 permission set, so
'trek-plugin-sdk validate' (and pack/publish, which run it) hard-rejected any
manifest using the newer scopes - db:write:reservations, notify:send,
hook:map-marker-provider and 25 more. Sync KNOWN_PERMISSIONS with the server
envelope (48 entries), mirror the full PluginContext (reservations,
accommodations, notify/ai/oauth/settings, packing writes + bags, file writes,
collab, tags/todos/daynotes/collections/atlas/vacay/journal, weather,
categories), type the tableContributor/mapMarkerProvider hooks + the
entity/entityId event hint, accept the day-detail widget slot, and extend
createMockHost so plugin unit tests can exercise all of it.

* feat(plugins): grant-scoped entity snapshots on core events

An events:subscribe handler so far learned only WHICH entity changed - useful
for cache busting, useless for reacting to content, and the userless handler
can't refetch. Now the broadcast tap derives a whitelisted field snapshot of
the changed entity and the supervisor attaches it per plugin, only where the
granted set holds the family's matching db:read:* permission (trips family ->
db:read:trips, budget -> db:read:costs, packing -> db:read:packing, dayNote ->
db:read:daynotes, file -> db:read:files). No acting user is ever synthesized.

The whitelists are explicit per family, so user ids (owner/paid_by/uploaded_by/
participants/members), trips.feed_token and future migration columns never
travel; a private packing item (#858) yields no snapshot at all because its
core broadcast is owner-scoped; deletes/reorders/bulk ops carry none.

* feat(plugins): pdf-section, atlas-layer and journal-entry provider hooks

Three more declarative provider surfaces in the map-marker mould - plugins
return data specs, the host normalizes, caps and renders; a slow or failing
provider contributes nothing:

- hook:pdf-section-provider: sections (title + paragraphs + a simple table)
  appended to the trip PDF export, escaped into the same HTML/print pipeline
  as the core content
- hook:atlas-layer-provider: per-user country tint layers on the Atlas map
  (ISO 3166-1 alpha-2 codes only, tone-whitelisted, non-interactive pane so
  mark/unmark clicks keep working)
- hook:journal-entry-provider: extra rows on a journal entry card, gated by
  the same journey access check as the journal routes + the Journey addon

Permission labels in all 22 locales, consent PERM_KEYS, SDK types + manifest
validator in both SDK copies, wiki tables, per-controller hardening tests.

* feat(plugins): trip-page plugins can replace core planner tabs and pick their spot

A trip-page plugin that takes over a core surface (a transit planner
superseding Transports, a costs plugin superseding the budget tab) had to sit
awkwardly next to the tab it replaces. capabilities.tripPage now names the
core tabs to hide while the plugin is active - whitelisted (transports,
buchungen, listen, finanzplan, dateien, collab), 'plan' deliberately not
replaceable, and the tabs return the moment the plugin is deactivated - plus
an optional 0-based position for the plugin's own tab. The feed re-validates
the values out of the DB blob so a hand-edited row can't hide anything else,
the admin list chips a replacing plugin (all 22 locales), and a saved session
tab that got replaced falls back to the plan view.

Also fixes the plugins feed dropping the day-detail widget slot to 'sidebar',
which would have mounted a day-panel widget on the dashboard.

* fix(plugins): audit follow-ups — normalization, secret cleanup, cron leak, slot filter

Adversarial audit of the whole plugin PR surfaced 12 confirmed issues; this
addresses them:

- place-details provider was the ONE hook controller with no normalization: a
  plugin's href/label/value went to the client raw and unbounded. Now normalized
  like journal-entry-rows (safeUrl http/https/mailto, length + count caps).
- trip-warnings capped message length + per-provider count (was unbounded).
- uninstall(deleteData) now also purges plugin_user_config, plugin_oauth_tokens,
  plugin_oauth_state, plugin_meta_migrations and the capability audit — encrypted
  per-user API keys + OAuth refresh tokens no longer survive a 'delete all data'
  and get silently re-adopted on a same-id reinstall.
- supervisor: a crash-restart cycle leaked the dead child's node-cron tasks and
  re-scheduled fresh ones, so a job fired N+1 times per tick after N crashes.
  onExit now stops them, mirroring kill().
- dashboard sidebar no longer mounts place-detail/day-detail widgets (they belong
  in the planner panels).
- reservation endpoint validation relaxed to match the 3.2.1 service: a coord-less
  endpoint is accepted and dropped downstream instead of BadParams (no breaking
  change), while a bad role/non-string still rejects up front.
- a replaced core tab reached by programmatic nav now falls back to the plan view.
- trips.update caps title/description like the places path; plugin-db guard bans
  load_extension as defense-in-depth.
- wiki: event snapshots, string-typed context ids, dayId in the payload, the live
  provider hooks and the costs update/delete grant are now documented correctly.

* feat(plugins): phase-0 lifecycle hardening + per-plugin RPC rate limit

Operational-readiness fixes from the completeness audit:

- Re-activation after a failure worked again: a plugin left in 'error' state by
  a load-failure or crash-auto-disable stayed in the running map, so the admin's
  'enable' button was a silent no-op. activate() now replaces a dead entry.
- Per-plugin RPC rate limit at the dispatch boundary: every ctx.* call runs
  synchronously on the host thread, so a plugin in a tight loop could freeze the
  whole instance (and the reap sweep). A token bucket (generous burst) + an
  in-flight cap now throttle a runaway plugin with a retryable HOST_ERROR; a
  legitimate plugin never notices.
- plugin_error_log retention (500 rows/plugin) so a crash-looper can't grow
  trek.db without bound; the crash-timestamp array is trimmed to its window too.
- TREK_PLUGIN_PERMISSIONS=off now logs a loud one-time warning that the OS
  permission jail is disabled.

* feat(plugins): read symmetry + broker — collab/journal/atlas reads, file content, trip create, rates

The plugin API leaned write-heavy: collab and journal could be written but not
read, files listed but not read, and there was no way to create a trip or see
exchange rates. This closes those gaps in the established RPC+gate pattern (zero
architecture risk), and it's what unlocks the importer + finance plugin classes:

- collab reads: ctx.collab.listNotes/listPolls/listMessages under a new
  db:read:collab (membership + Collab addon, like the REST GETs)
- ctx.journal.getEntries(journeyId): a journey's entries, journey-access-checked,
  under the existing db:read:journal
- ctx.atlas.bucketList(): the acting user's bucket list, under db:read:atlas
- ctx.files.getContent(tripId, fileId): a file's bytes as base64 under a NEW
  db:read:files:content grant (reading a passport scan is more sensitive than its
  filename), size-capped at 10MB before it crosses the IPC pipe, trashed files
  refused
- ctx.trips.create(input): a new trip owned by the acting user, gated by the app's
  trip_create right + a bound user — the capability importers need
- ctx.rates.get(base): cached currency exchange rates, tenant-free like weather

Also caps trips.update title/description like the places path, and the plugin-db
guard now bans load_extension (defense-in-depth). SDK, mock-host, i18n (22
locales), consent labels and the wikis are all in lockstep.

* feat(plugins): deeper integration + user-facing activity transparency

Wave 2 of the completeness work — richer extension points, deeper metadata, and
the transparency that makes the broad read grants accountable:

- db:meta now attaches to reservations + accommodations too (not just
  trip/place/day), gated by reservation_edit / day_edit respectively — the
  natural home for an external-id mapping (AirTrail/calendar/booking-import sync)
  without forking the core schema.
- reservation-detail widget slot: a widget can mount on a booking card, scoped to
  the open reservation via reservationId in trek:context (the place-detail /
  day-detail pattern, third instance).
- tableContributor gains the transports + todos views, so a plugin can add
  host-rendered columns/actions there too.
- User activity log: GET /api/plugin-activity + a Settings → Plugins panel showing
  every host-mediated action a plugin took bound to the signed-in user, across all
  plugins, newest first — the user-facing half of the hash-chained audit. This is
  what legitimizes the deliberately broad read grants: not just the admin, the
  person whose data is read can see what was done in their name.
- DX: the local dev server now binds a default acting user, so the canonical
  ctx.trips.getPlaces(tripId) call works locally instead of failing RESOURCE_
  FORBIDDEN; the create scaffold drops the dead manifest routes[] / capabilities.nav
  fields the host ignores.

SDK, i18n (22 locales), consent labels and the wikis are all in lockstep.

* fix(memories): load Immich album photos on Immich v3

Immich v3 removed the `assets` property from AlbumResponseDto, so
`GET /api/albums/:id` no longer carries album contents. TREK read album
photos from that property, which now parses as undefined and degrades to
an empty array — hence "No photos yet" in the Journey gallery picker even
though the album header shows the right count (that count comes from
`GET /api/albums` -> assetCount, which v3 still returns).

Two call sites read the removed property. Besides getAlbumPhotos (the
reported bug), syncAlbumAssets failed silently on v3: it reported
`success: true, added: 0` while syncing nothing.

Fetch album contents via an `albumIds`-filtered `POST /api/search/metadata`
when `assets` is absent, and feature-detect rather than probe a version.
The two paths are not interchangeable: on v2, searchMetadata
unconditionally scopes results to `[self, ...partners]`
(`asset.ownerId = ANY(userIds)`), so an albumIds search against an album
shared by a non-partner returns nothing. v3 added an albumIds branch that
checks AlbumRead and skips that owner filter. v2 also hard-defaults
`visibility` to `timeline`, dropping archived assets. So v2 must keep
reading the album detail body, which this preserves exactly.

`withExif: true` is required on the search path: it has no default and
gates an inner join, so without it Immich omits `exifInfo` entirely and
every photo's city/country goes null.

The existing test mock returned an album detail body *with* `assets` — it
encoded the v2 assumption, which is why this shipped green. It now models
v3 by default, with explicit v2 coverage asserting no search call is made.

Fixes #1492

* feat(plugins): daily AI/notify budgets, runtime scheduler & reliable event redelivery

Per-plugin daily caps on ai.complete/ai.extract and notify.send (defaults
200 / 100, overridable via TREK_PLUGIN_AI_PER_DAY / TREK_PLUGIN_NOTIFY_PER_DAY),
seeded from the capability audit so a mid-day restart resumes the count instead
of resetting it. Surfaced at GET /plugins/:id/budget.

ctx.scheduler (at / in / every / cancel): persistent, userless timers that
survive restarts and fire a scheduled() handler, riding the existing jobs:run
grant so no new consent or admin setup is needed. Backed by
plugin_scheduled_tasks, swept every 30s, capped at 100 tasks/plugin with an 8 KB
payload and a 60s recurring floor; rows are removed on uninstall.

Core events that fire while a subscriber is mid-restart are now held in a
bounded in-memory buffer (200/plugin, 15 min TTL) and replayed once it goes
active again, with the events:subscribe grant and snapshot gating re-evaluated
at replay time so nothing leaks if a grant was revoked while the plugin was down.

* feat(plugins): GDPR data-subject rights — durable per-plugin erasure + export

New hook:user-data grant with two userless lifecycle handlers a plugin can put
on its definition: deleteUserData and exportUserData. Neither carries an acting
user — the plugin only learns the userId and touches its own db — so the grant
reads nothing from core data; it exists purely so a plugin can honour a GDPR
erasure or data-access request.

When a TREK account is deleted (admin or self-service), every installed plugin
holding the grant gets a row in a new durable erasure queue and its
deleteUserData runs on the next sweep, retried until it ACKs — so erasure
survives the plugin being offline or the server restarting. The core deletion
path notifies the runtime through a dependency-free relay (like the event sink),
keeping the auth/admin services decoupled from the plugins layer, and a plugin
bookkeeping error can never fail the account deletion.

Portability is served by GET /api/admin/plugins/user-data/:userId/export, which
fans exportUserData out to the active granted plugins and aggregates what each
holds about the user. Queue rows are purged on uninstall; the grant is labelled
in all 22 locales.

* feat(plugins): atomic ctx.db.tx for consistent multi-write on a plugin's own db

Plugins could already query/exec/migrate their own SQLite file, but a multi-step
write (move an item between tables, decrement one row and increment another) had
no way to be atomic. db.tx([{sql, args?}, …]) runs up to 100 statements in a
single transaction — all commit or all roll back — and reads within the batch see
its own earlier writes, so read-modify-write is safe. Each op is one statement:
a read returns { rows }, a write { changes }. The same guard (no ATTACH/PRAGMA/
RECURSIVE, size + row caps) applies to every statement in the batch.

* fix(memories): filter hidden Immich assets at the source, not just the picker

#1474 has the same root cause as #1492: the Immich v3 migration. On v2,
searchAssetBuilder hard-defaulted metadata search to `timeline` visibility
(`visibility = options.visibility ?? Timeline`), so hidden Live Photo
motion parts could never come back from a search. v3 defaults to any
visibility except `locked`, so they do — which is why the reporter is on
Immich 3.0.1 and why the bug never appeared before.

Ask for `visibility: 'timeline'` explicitly on the search path. That
restores v2 semantics on both versions and stops hidden assets crossing
the wire, which also fixes a pagination wart: a full page half-made of
motion parts previously rendered as a half-empty page, because hasMore
counts the raw page length while the filter shrinks the rendered set.

The client-side filter was display-only, applied in searchPhotos and
getAlbumPhotos — both picker-listing paths. Nothing guarded persistence
or rendering: getOrCreateTrekPhoto stores any id it is handed, pipeAsset
forwards Immich's 400/404 verbatim, and the photo grid is a plain <img>
with no onError. So syncAlbumAssets, which filtered `type === 'IMAGE'`
only, could persist a hidden IMAGE as a permanently broken tile. It now
applies the same guard, extracted as isVisibleAsset().

Albums keep their filter rather than requesting `timeline` visibility:
albums legitimately contain archived assets, and both the v2 album body
and the v3 album search return them.

Does not address tiles already persisted before this — those still render
broken and need a separate fix.

Refs #1474

* docs(memories): correct Immich version boundaries in the hidden-asset comments

Verified against the v1.120.0 → v3.0.0 OpenAPI specs and server source. The
previous comments said "Immich v2 hard-defaulted metadata search to timeline
visibility". That is true only for 1.133–1.144.

- `visibility` was added in 1.133.0. Before that, searchAssetBuilder applied
  `.$if(options.isVisible !== undefined, ...)` with no default, so pre-1.133
  servers returned hidden assets too. #1474 was therefore not purely a v3
  regression.
- Those servers strip the `visibility: 'timeline'` filter rather than
  rejecting it: Immich validates with `whitelist: true` and no
  `forbidNonWhitelisted`. So the request stays valid, the filter is a no-op,
  and isVisibleAsset() is the ONLY guard there. Say so, so it does not get
  removed later as redundant.
- `albumIds` only exists from 1.135.0. Because unknown properties are stripped,
  an albumIds search against an older server would silently drop the album
  filter and return the entire library as the album's contents. Feature
  detection on `assets` (present through 1.144.1, absent on v3) makes that
  unreachable; a version probe with a wrong boundary would not.

Also cite Immich's own enum, which documents AssetVisibility.Hidden as
"Video part of the LivePhotos and MotionPhotos".

Comments only — no behavior change.

* feat(plugins): dashboard trip-card badges + a mock-host driver for plugin tests

Two additions that round out the plugin platform's breadth and its authoring DX.

tripCardProvider hook (hook:trip-card-provider): a plugin returns small declarative
badges for the dashboard trip cards. The dashboard fetches all visible cards in one
call; the host access-checks every tripId for the acting user, bounds each field
(label/value length, enum tone, http/https/mailto-only url), caps the count and drops
any badge for a card that wasn't requested — plugin JS never runs on the dashboard.
Rendered as text chips under the card meta; labelled + gated in all 22 locales.

createMockHost now exposes run(def) — the other half of a plugin unit test. Where the
ctx recorders capture what a plugin read, run() fires its own entry points (route, job,
scheduled, event, plugin-event, deleteUserData, exportUserData, provider hooks) against
the same mock ctx, and host.scheduled surfaces the timers it armed. A handler the plugin
didn't declare throws a clear error instead of a silent no-op.

* feat(plugins): include plugin data + code in backups, applied on restart

A TREK backup archived travel.db + uploads + the encryption key, but each plugin's
own SQLite file — the ONLY copy of the user data it holds — and its installed code
lived in separate trees that were never captured, so a restore left the plugins rows
with no data or code behind them.

createBackup now adds plugins-data/ (each plugin's db + WAL sidecars, so SQLite
recovers a consistent snapshot) and plugins-code/ (skipping dev-links by realpath, so
an author's linked source is never bundled). Restore can't swap those live — the
runtime holds each plugin db open — so it STAGES the extracted trees beside the live
ones and the runtime swaps them in at the next boot, before it opens anything. Same
"applies on restart" model the bundled encryption key already uses: no plugin quiesce,
no swap under open handles, no new admin setup. Older archives without the trees restore
exactly as before.

* fix(plugins): audit — runtime robustness, security & data-lifecycle fixes

Fixes from an adversarial audit of the plugin system, host/runtime side:

Robustness:
- getPluginDataDb recreated a handle a terminal-failure dispose had closed but
  left cached, so a re-enabled plugin's db:own threw on every call — recreate
  when the cached handle is shut.
- ctx.ws.broadcast* now carry _inv, so the host can bind the acting user (the
  capability was silently refused, i.e. dead, without it).
- ctx.events.emit swallows a rejected emit instead of crashing the child into a
  terminal 'error'; an uncaught throw AFTER activation is treated as a crash
  (restart with backoff), not a load failure.
- A crash-respawned child gets the same activation deadline as a first activation
  and the buffered-event queue is cleared on the timeout path, so a hung onLoad
  after a crash can't peg a core and orphan events forever.
- Expired buffered events are pruned by the reaper, not only at flush; the
  scheduler + erasure sweeps scope their LIMIT window to ACTIVE plugins so a
  backlog for inactive plugins can't starve deliverable work.

Security / integrity:
- Unix-domain-socket / named-pipe connects are refused by default in the egress
  guard (a host-local pivot to docker.sock / DB sockets), under the same policy
  as private IPs.
- db.tx refuses transaction-control statements (a raw COMMIT would break its
  atomicity) and caps rows across the WHOLE batch, not per statement.
- plugin_capability_audit is retention-capped per plugin (chain-safe: retained
  rows stay self-verifying), so it can't grow unbounded in the shared db.
- A cap of 0 in TREK_PLUGIN_AI_PER_DAY / _NOTIFY_PER_DAY now disables the broker
  instead of falling back to the default.

GDPR data lifecycle:
- Account deletion now erases host-side per-user plugin tables (config, OAuth
  tokens/state) and enqueues the own-db erasure from the CORE path, so it works
  even when the runtime is disabled or pre-boot; guest deletion does the same.
- uninstall keeps a pending erasure when data is retained (deleteData=false);
  erasure delivery is no longer grant-re-checked (a queued erasure is a duty);
  export flags installed-but-inactive plugins as pending instead of omitting them.

Backup/restore:
- Plugin DBs are WAL-checkpointed before archiving (no torn/stale snapshots).
- Restore applies the staged trees immediately by quiescing the plugins (no
  unbounded gap where a later unrelated restart would revert diverged data);
  the swap is content-level (safe on a volume-mounted root) and preserves
  dev-links; the decompressed-size cap is operator-raisable.

* fix(plugins): audit — hook-output hardening, dashboard slot & mock-host parity

- Map-marker and atlas-layer tones were validated on String(tone) but emitted
  raw, so a non-string tone (an object with a matching toString) slipped through
  and crashed the client that renders it — check the raw value against the enum.
- View-contribution column/action caps are now PER ENTITY, not per view, so a
  plugin's columns no longer vanish from every table row past the first 20; the
  dashboard trip-card badge cap is per card (≥ one on every visible card).
- A reservation-detail widget no longer also renders as a context-free dashboard
  sidebar card (the inline filter was missing that slot).
- mock-host matches the real host: it ignores asUserId on trip reads (bind the
  acting user), throws on a wrong user-scope notify target instead of coercing,
  enforces the scheduler caps, and detects RETURNING as a read in db.tx — so a
  passing author test can't hide a production RESOURCE_FORBIDDEN.

* feat(plugins): full ctx parity in the dev server + fire jobs/events/hooks locally

The trek-plugin dev server injected only ~6 of the ~35 ctx areas, so any plugin
touching ctx.costs/packing/files/notify/ai/settings/scheduler/meta/oauth/db.tx/…
hit a TypeError in local dev while the same code passed mock-host tests and worked
installed. It also could only exercise routes.

Delegate every non-db-own capability to a grant-enforcing mock host (the same one
unit tests use) while keeping the real node:sqlite for db:own and dev-native ws
capture + logging — so the whole surface works in dev with the exact production
permission rules. dev-fixtures.json now takes the createMockHost options shape, so
you can seed the full surface. New GET /__dev/fire/<kind>[/<name>][/<fn>] fires a
job, scheduled timer, event subscription, GDPR handler or provider hook against the
dev ctx, closing the "can't test non-routes locally" gap.

* feat(plugins): wire the photoProvider + calendarSource hooks to real core consumers

Both hooks were declared, typed and documented but NO core code ever invoked them,
so an author could build, mock-test and install a photo or calendar plugin that
silently did nothing. Give each a real consumer that fans out to it, exactly like
the other eight provider hooks:

- GET /api/plugin-photos/search (+ /sources, /item) aggregates photoProvider results
  for the picker — {id, title?, thumbnailUrl, fullUrl, takenAt?}, thumbnail/full URLs
  http/https-only (they become <img src>), per-source count capped, failing source
  skipped.
- GET /api/plugin-calendar?start=&end= aggregates calendarSource events for the
  signed-in user — {id, title, start, end, allDay} ISO, count capped, failing source
  skipped, sensible default window.

Both run with the acting user bound. The SDK interfaces now pass ctx as the last arg
(so a source can reach ctx.settings/oauth/http for its backend), and the wiki marks
them live instead of "reserved — no core consumer".

* feat(plugins): close the create-heavy API asymmetries importers/sync hit

Core services implemented these but plugins had no path to them, so the flagship
importer/sync integrations hit real walls. Added, each reusing the EXISTING grant
(no new consent):

- ctx.trips.removeMember(tripId, userId) — reconcile DEPARTURES, not just additions
  (db:write:members + member_manage). Never removes the owner (that would orphan the
  trip); ownership transfer stays a separate deliberate action.
- ctx.journal.createJourney({title, subtitle?, trip_ids?}) / deleteJourney(journeyId)
  — an importer can now bootstrap the journal it fills with entries and clean it up
  (db:write:journal), instead of only appending to journals a human created first.

Wired end-to-end (envelope → rpc-host → create-rpc-host reusing tripService/
journeyService → both SDK copies → mock-host) and documented. (trips.delete needs its
own destructive permission + consent copy and collab edit/delete + collections.delete
remain — tracked as small follow-ups.)

* feat(plugins): strip emojis from plugin-rendered text so it matches TREK's lucide UI

Plugin authors (especially AI-generated ones) sprinkle emojis into the declarative
text TREK renders in its OWN chrome — hook contributions (badges, columns, warnings,
PDF sections, map-marker/atlas labels, journal rows, place details, trip-card badges,
calendar + photo titles) and notifications — which clashes with TREK's lucide-only icon
language.

A shared stripEmoji() removes emojis (incl. flag/ZWJ/variation-selector sequences) and
tidies the leftover whitespace, applied at the render boundary in every hook-contribution
normalizer and in notify.send — so no matter what a plugin returns, the text TREK draws
stays emoji-free. It does NOT touch a plugin's own sandboxed /ui frame (the author's to
design), and it leaves photo ids verbatim (they round-trip to getById). The validate CLI
warns when a manifest name/description contains emojis, nudging authors to the declarative
`icon` field (a lucide name) instead.

* fix(plugins): harden the restore-apply path — regressions from the backup/dev fix pass

A final audit of the fix pass caught three regressions clustered in the two newest
surfaces; the restore path could both crash the server and destroy data.

- CRITICAL: a restore quiesces plugins via supervisor.shutdownAll() AFTER closeDb(), but
  shutdownAll killed children without first marking them stopped, so each child 'exit'
  took the CRASH path and wrote crash-accounting rows into the now-closed core DB — the
  throw escaped an EventEmitter listener as an uncaughtException and killed the whole
  process mid-restore. shutdownAll now marks every entry stopped and drops it from
  `running` BEFORE the kills (so onExit early-returns), and the onStatus/onLog DB hooks
  are wrapped in try/catch (also covers the stderr→onLog path). This also stops a normal
  shutdown from logging phantom "crashed" rows.
- HIGH: swapContents cleared live entries then MOVED staged ones in, so a crash mid-move
  permanently deleted a plugin's only data copy (staging was already emptied, so a retry
  couldn't restore it). It now COPIES each staged entry over the live one and only deletes
  staging at the very end — `staged` stays the complete source of truth, making the whole
  operation crash-idempotent.
- HIGH: the dev server lost the actingUserId=1 default in the mock-host refactor, so a
  fresh scaffold refused every user-bound capability. Restored.

* fix(plugins): final-audit medium/low findings

- GDPR export flags an active plugin whose export errored/timed out as `pending`
  instead of silently omitting it (collectUserExport now returns a discriminated
  result), so a data-access export never reads complete while missing data.
- Account deletion also enqueues an erasure for plugins UNINSTALLED with retained
  data (an orphan data dir) — a same-id reinstall now honours the deletion instead
  of re-adopting the user's data forever.
- oauth.getToken returns null in a userless context (matching the SDK/mock contract)
  instead of throwing RESOURCE_FORBIDDEN a background caller can't handle.
- Crash-backoff restart is identity-guarded (+ the timer is tracked and cleared like
  the activation timer), so a disable + re-enable during the backoff window can no
  longer respawn a ghost child from the replaced entry.
- db.tx transaction-control guard strips leading comments first, so `/* */COMMIT`
  can't slip past the start-anchored check and break batch atomicity.
- createJournal inherits its cover only from a trip that was actually LINKED
  (access-checked), closing a cross-tenant cover-image read on plugin + REST paths.
- trip-warnings drops a null array element instead of losing ALL of that provider's
  warnings; plugin-activity floors a non-integer ?limit so it can't 500.
- The trek-plugin dev server binds loopback only and refuses cross-site requests to
  its side-effectful /__dev/fire endpoints (it serves real routes + no-auth dev
  actions).

* fix(plugins): clear no-misleading-character-class in the emoji stripper

The character class listed the ZWJ, variation selectors and combining keycap
marks as members, which eslint reads as an accidental combined grapheme and
rejected on CI. Pull the emoji glyphs out into Extended_Pictographic /
Regional_Indicator alternatives so only the joiner/selector code points stay in
the class, with a scoped disable where the rule still can't tell them apart.
While here, reset lastIndex before the /g regex is reused in hasEmoji() so a
second call can't resume mid-string and miss a leading emoji.

* fix(security): trip-scope note-file deletion and guard the LLM base URL

Two reported issues:

- deleteNoteFile only matched on the note id and file id, so a member of trip A
  could delete a file attached to a note in trip B by guessing its id. Thread the
  trip id through the service and controller and scope the delete to it, the way
  every other collab operation already does.

- The LLM extraction clients fetched the user-configured base URL directly, so a
  user could point it at the cloud-metadata endpoint (169.254.169.254) and read
  the echoed error body. Route both clients through a new safeFetchLlm() that
  blocks the link-local/metadata range while still allowing a local or LAN Ollama
  (loopback and private ranges stay reachable), pinned to the resolved IP so a
  hostname can't rebind to the metadata address after the check.

* fix(security): route every LLM client through the SSRF guard

The base-URL SSRF fix covered the openai-compatible and anthropic clients but
missed the native Ollama /api/chat client and the /api/tags + /api/pull model-
management calls, which fetched the same user/admin-configured URL raw. Route
them through safeFetchLlm too, and harden its link-local check: cover the whole
fe80::/10 range (not just the fe80: prefix), the IPv4-mapped/-compatible metadata
spellings, and the AWS IMDSv6 endpoint, while still allowing a loopback or LAN
model server.

* fix(plugins): narrow emoji stripping to real emoji + close two gaps

The stripper used \p{Extended_Pictographic}, which also swallowed legitimate text
symbols a plugin renders — (c), (r), tm, rating stars (a full row collapsed to the
empty ones). Narrow it to \p{Emoji_Presentation} (+ skin-tone modifiers, regional
indicators, and the ZWJ / variation-selector / keycap / tag glue) so only colour
emoji go and text symbols stay. The whitespace tidy-up now runs only when
something was actually removed and never collapses newlines, so plain text and
PDF paragraphs keep their exact spacing.

Also strip the trip-warning message — the one render boundary that slipped
through — and skip stray files (not just directories) when enqueuing orphan
plugin-data erasures.

* fix(plugins): harden the supervisor lifecycle + restore staging

- Restore staging is now atomic: the plugin trees are copied into a `.tmp` sibling
  and only a fully-copied tree is renamed to `.restore`. A copy that died partway
  (disk full, crash) previously left a PARTIAL `.restore`, and the next boot's swap
  deletes every live plugin dir missing from it — data loss, after the restore had
  reported success.
- A late `loaded` / `load-error` from a child that finished during the kill grace no
  longer resurrects a stopped/removed plugin or re-registers routes on an active one
  (which leaked node-cron tasks and double-fired jobs after a re-enable).
- The crash-backoff branch clears the stale activation deadline so it can't fire
  during the wait and cancel the scheduled retry.
- shutdownAll rejects in-flight activations (no more hung enable requests) and clears
  the buffered-event queue (no replaying pre-restore events); activate() self-handles
  the rejection so a caller that doesn't await can't crash the process.
- Erasure drains coalesce onto the one in flight instead of running concurrently and
  delivering an erasure to a plugin twice.

* fix(plugins): consistent plugin-db backups + close two restore/handle races

- Backups archive a VACUUM INTO snapshot of each plugin db instead of the live file.
  The archiver reads lazily while streaming, so a plugin writing during the backup
  could land a torn .db plus an out-of-sync -wal in the zip — the plugin's only data
  copy, silently corrupt. The snapshot is point-in-time consistent and drops the
  sidecars.
- The host data handle is resolved lazily per call rather than captured once, so a
  disable+re-enable within the kill grace can't have the old host's dispose() close
  the db out from under the freshly re-activated one.
- A restore quiesces the running plugins regardless of whether the archive carried
  plugin trees. An older archive (no trees) swaps in a different `plugins` table while
  the live children keep running with their pre-restore identity — ghosts, invisible
  in the restored UI and unstoppable short of a restart.

* fix(plugin-sdk): bring the mock host and dev runner to parity with the real host

A plugin developed against the local mock/dev runner must behave the same as in
production, or authors debug phantoms. Sixteen behaviour drifts closed:

- Userless handlers (jobs, scheduled tasks, events, GDPR erasure/export) run with a
  userless ctx like the real host, so membership-checked reads correctly refuse.
- member add/remove: member_manage gate, owner-guard, target-exists check.
- journal: addon gate + entry_date/title validation + access checks.
- notify: emoji stripping, title/body caps, in-app link check.
- scheduler dueAt must be finite and within a year.
- db.tx: TX_CONTROL / FORBIDDEN / op-cap in both the mock and the dev runner.
- ws/users: acting-user, membership, self-only and canSeeUser gates + field whitelist.
- dev db SQL guards + row cap; the dev request always carries headers; dev user is
  no longer implicitly an admin.
- core-write and small field validations mirrored from the shared schemas.
- addon toggles + a generic per-trip permission record; declaredEmits warning.
- /__dev/fire cross-site guard extended to cover the /api routes.

* fix(oauth): don't revoke active MCP sessions on a normal token refresh

refreshTokens() unconditionally revoked all MCP sessions for the
(user, client) pair on every successful refresh-token grant, not only
on replay/theft detection. Since MCP access tokens expire hourly, any
long-lived MCP client (Claude Code, Claude Desktop, etc.) silently
refreshing its token in the background would have its live session
killed roughly every hour, causing /mcp to return 404 "Session not
found" until the client was manually reconnected.

Session revocation now only happens on the replay-detection branch,
which is the actual security-relevant case; mcpHandler already
re-validates session.userId/clientId against the current token on
every request, so a legitimate rotation doesn't need to pre-emptively
kill the session.

Fixes #1475

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* fix(atlas): match region names regardless of accent/dash formatting

The region layer's name-matching fallback (used when the geocoder's cached
region_code doesn't match the bundled admin-1 code) compared raw strings, so
a cached region_name with different accenting or dash style than the bundle
(e.g. "Ile-de-France" vs "Île-de-France") silently failed to highlight.
Added normalizeRegionName (diacritic-insensitive, dash-insensitive) and used
it everywhere region names are compared.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* fix(atlas): connecting-flight layovers no longer mark the layover country visited

getStats' #1366 transport-only-country merge read every reservation_endpoints
row for a trip, including role='stop' (a connecting-flight layover). A plane
change through a country's airport now no longer marks it visited — only
real from/to legs do.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* feat(i18n): improve Simplified Chinese translations

* fix(i18n): correct Vietnamese locale translations (#1502)

* fix(pwa): self-host the leaflet stylesheet instead of loading it from unpkg (#1497)

The service worker's cdn-libs rule cached the unpkg leaflet.css as an
opaque response, which the browser then rejected for the CORS <link>
request — Atlas and trip maps stayed blank until users manually
unregistered the service worker. Bundling the stylesheet from the
leaflet package removes the cross-origin fetch (and the runtime CDN
dependency) entirely, so it is precached with the rest of the app
shell. The now-dead unpkg runtime-caching rule is dropped.

* fix(packing): select quantity on focus so typed number replaces it (#1513)

* fix(planner): sort transport and booking cards chronologically (#1507)

* fix(planner): keep leg distance visible across a car rental's middle days (#1504)

* fix(costs): restore the expenses CSV export lost in the Costs rework (#1500)

* fix(client): disable request timeout for cover image uploads (#1495)

* fix(planner): scope transit quick picks to the selected day (#1460)

* fix(atlas): restore the missing Spain bounding box so southern-Spain points stop resolving to Algeria (#1490)

* fix(mobile): dvh shell, safe-area padding in Plan/Places drawer, packing name overflow (#1505)

* fix(planner): persist typed booking addresses and add autocomplete (#1496)

* fix(security): enforce trip read scope and share flags server-side

get_trip_summary returned the trip's members (with emails), full itinerary
and accommodations to any valid MCP token regardless of its OAuth scopes.
Gate the core bucket on trips:read, keeping id + title so the tool still
works for navigation (GHSA-qvw8-w937-vcmq).

The public trip share ignored share_map: days, assignments, notes and the
place list (with coordinates) were always returned. Gate them on share_map
server-side, mirroring the journey share. Same class nearby: the journey
photo/asset proxies now honour share_gallery and the shared place-photo
proxy honours share_map.

* test(dashboard): pin the clock so trip-window fixtures don't rot

DashboardPage's tests seed a trip dated 2026-07-01..07-10 and assume it is
the current/spotlight trip. Once the wall clock passes that window the
past/upcoming split relocates the trip and 14 tests fail with "Paris
Adventure" no longer rendered. Freeze "now" to a date inside the window
(shouldAdvanceTime keeps userEvent/waitFor working) so the suite is
deterministic regardless of when it runs.

* fix(atlas): derive country boxes from admin0 and make country removal stick (#1490, #1486)

Removing a visited country in Atlas was a no-op, and 43 countries could not be
geocoded at all. Two independent defects behind one report.

The bounding boxes were a hand-maintained table that had drifted from the admin0
polygons we already ship: NG, BY, GL, KP, TD, SS and 37 others had no box, so
their coordinates fell into a *neighbour's* box as the sole candidate — Lagos
resolved to Benin, Minsk to Russia. getCountryFromCoords also short-circuited a
lone bbox candidate without checking the polygon, which is how a point in
southern Spain became Algeria. Derive the boxes from admin0 instead, one box per
geometry part so countries straddling the antimeridian (RU, US, FJ, KI) don't get
a globe-spanning box, and always validate against the polygon.

Removal never persisted because Atlas re-derives countries from places and
transport endpoints on every request: those have no visited_countries row, so the
DELETE hit nothing, the client hid the country optimistically, and the next
getStats brought it straight back. Tombstone the removal in a new
hidden_countries table and suppress it in the derived passes. Re-marking lifts
the tombstone; adding a real place there brings the country back on its own.

The same #1366 endpoint merge had been copy-pasted into authService.getTravelStats
without the role filter Atlas got, so the dashboard passport card still counted a
plane-change country as visited. Filter it there too, and honour the tombstones so
the card and the map agree.

CONTINENT_MAP had the mirror-image gap (ES missing, NG present) and bucketed the
newly-resolvable countries into 'Other'; fill it out to cover all 198.

* fix(ics): compute all-day DTEND without a local-timezone round-trip (#1453)

exportICS built all-day DTENDs as `new Date(date + 'T00:00:00')` → setDate(+1) →
toISOString(). The string carries no Z, so it parses as *server-local* midnight and
the toISOString() round-trip lands a day early on any positive UTC offset. DTEND is
exclusive, so every trip exported from a server east of Greenwich silently lost its
last day — and every per-day summary event lost its only day.

Reuse dayService.addDays(), which stays in Date.UTC throughout, for both blocks.

This shipped green because the all-day test asserted DTSTART and SUMMARY but never
DTEND, and TZ is pinned nowhere in the server test setup — so CI inherited UTC, one
of the few offsets where the bug is invisible. The new test pins TZ across
Europe/Berlin, Asia/Tokyo, Pacific/Kiritimati, America/New_York and UTC; it fails
against the old code in exactly the three positive-offset zones.

* fix(client): route every multipart upload through a timeout-free helper (#1495)

The shared axios instance carries timeout: 8000, and axios' timeout is a whole-request
deadline rather than an idle one — so any upload whose body takes more than 8s to push
is aborted mid-stream, which the server surfaces as a multer "Request aborted".

#1495 fixed this by hand-writing `timeout: 0` on the three cover uploads. That left the
identical bug live on 7 of the 15 multipart call sites, including the two that accept
500 MB: the documents addon (filesApi.upload) and backup restore. Avatar uploads — a
phone photo on a slow uplink, exactly the reported scenario — were also still on 8s.

Add postMultipart() and route all 15 sites through it, so opting out of the timeout is
the default instead of something each new endpoint has to remember. It also collapses
the three different hand-rolled Content-Type spellings; axios unsets the header for
FormData in the browser anyway so the platform can generate the boundary.

coverUpload.test.ts is superseded by uploadTimeout.test.ts, which pins all 13 endpoints
rather than just the three covers.

* fix(client): shared NumericInput so a typed digit replaces instead of appending (#1513)

Tapping a pre-populated numeric field parks the caret at the end, so the first digit
typed is appended rather than replacing: a packing quantity of 1 plus "6" committed 16.

#1513 fixed exactly one field — the packing quantity input. The weight field in the
very same row still appended, as did nine others, including Vacay's entitlement-days
which autoFocuses straight into the bug.

Extract the behaviour into components/shared/NumericInput and adopt it at all eleven
sites. It owns only the part that was uniformly broken — sanitizing and select-on-focus;
styling and commit semantics stay with each caller, since some save on every keystroke
and others on blur.

Two details worth keeping:

- The select() runs synchronously *and* on the next frame, because WebKit undoes a
  synchronous select() with its own post-focus caret placement. The deferred pass is
  guarded: if a keystroke lands inside that one-frame window, re-selecting would swallow
  it — "48.853" typed into a coordinate field came out as ".853" — so the first input
  cancels the pending select.
- The fields move from type=number to type=text + inputMode. Same keypad, minus the
  scroll-wheel-silently-changes-the-value and spinner-arrow footguns.

* fix(i18n): name Costs consistently across the app and in every locale (#1464)

The Budget addon was renamed to Costs, but the rename only ever landed on the trip tab
and — in English — the admin catalog. Two problems fell out of that.

The zh locale was self-contradicting: the admin add-on card read the English literal
'Costs' while the tab it configures read '费用'. That is the bug as reported.

Underneath it, the rename had been done in `en` and copy-pasted verbatim into the other
21 locales, so 16 of them showed an untranslated English "Costs" for the trip tab.
Meanwhile the surfaces the rename missed — the public shared-trip tab, its totals label,
the login feature list ("Budget Tracking") and the member-permission checkbox — were
still correctly translated as *Budget*. Every locale therefore called the same thing two
different things.

Translate "Costs" properly in all 22 locales and apply it to all six keys, so each
locale is internally consistent and none is left holding an English literal.

Note that i18n parity cannot catch this class of drift: it compares key and file sets,
never values, which is exactly why the zh mismatch shipped green.

* fix(i18n): call packing and todo categories "lists" (#1505)

#1505 asked for the "Add category" button to read "Add list". Renaming only the button
would have left it sitting next to "Delete Category" and "Change Category", and clicking
it would still open a field placeheld "Category name (e.g. Clothing)" — so the whole
user-facing vocabulary moves together, in all 22 locales.

Covers the button, the name placeholder and its inline variant, the delete and move menu
items, the delete-confirm dialog, and the todo sidebar heading, detail label and
no-category states.

The key names are unchanged, so i18n parity is unaffected — only values move.

The bulk-import hint is deliberately left alone: it documents the CSV column layout, and
the column really is `category` in the schema. Its strings have also drifted between
locales (several describe an entirely different format), which is worth a separate pass.

* fix(plugins): audit hardening — OAuth SSRF, demo upload guard, SDK perms, ICS tz

Follow-ups from the 3.2.2 audit, all additive and non-breaking:

- The host-side OAuth token exchange now runs through the SSRF guard
  (safeFetchLlm): a token_url that resolves to the cloud-metadata /
  link-local range is refused and the connection is pinned to the resolved
  IP, so a DNS name or IPv6 literal pointing at metadata can't be reached
  and can't rebind. Loopback and LAN stay reachable so a self-hosted
  internal IdP keeps working.
- Plugin files.create honours the demo-mode upload block the REST path
  enforces, so a demo user can't write files to a public demo instance
  through db:write:files.
- The SDK manifest validator learns hook:trip-card-provider and
  hook:user-data, which the host already accepts — it was hard-rejecting
  valid manifests that use them.
- ICS export validates a stored/plugin-provided timezone before it reaches
  Intl, so an invalid zone degrades to a floating local time instead of
  throwing a RangeError and crashing the calendar export / all-trips feed.

* fix(plugins): audit follow-ups — WAL backup, async read, dev-link gate, webhook raw body

More non-breaking hardening from the 3.2.2 audit:

- Backups copy a closed plugin.db's -wal/-shm sidecars. A plugin with no
  open handle has no writer, so the sidecars are a consistent set with the
  .db; without them an unclean shutdown's committed-but-uncheckpointed WAL
  rows would be missing from the backup. The open-handle path still captures
  a consistent VACUUM INTO snapshot and skips the sidecars.
- files.getContent reads off the event loop (fs.promises.readFile) instead
  of a synchronous up-to-10MB read + base64 that stalled every other plugin
  RPC and request for its duration.
- Plugin discovery only follows dev-link symlinks when TREK_PLUGINS_DEV_LINK
  is set, so a stale link left on the volume is not registered on a normal
  (non-dev) boot.
- Webhook (auth:false) proxy routes now receive req.rawBody, so a plugin can
  verify a provider's HMAC signature over the exact payload it was computed
  on. Enabled app-wide via Nest's rawBody option; authenticated routes never
  receive the raw bytes.

* fix(build): move tz-lookup to server production dependencies

tripService imports tz-lookup at runtime (ICS VTIMEZONE resolution, #1453)
but it was declared in devDependencies. Local dev worked because the
monorepo hoists devDependencies into the root node_modules, but the
production Docker stage installs with `npm ci --omit=dev`, so the image
shipped without tz-lookup and the server crash-looped with
MODULE_NOT_FOUND on boot. Move it to dependencies and sync the lockfile.

* fix(dashboard): round the boarding-pass wrapper so its mask layer isn't square

The #1459 Safari fix split the pass into an outer wrapper (left notch) and the
inner glass panel (right notch). The wrapper carries a mask + translateZ, which
promotes it to its own compositing layer whose bounds stayed square, showing a
square edge over the rounded inner panel. Give the wrapper the same 24px radius.

* fix(plugins): refresh the active-plugin store on admin changes

Activating/deactivating a plugin only updated the admin panel's own rows, not the
app-wide active-plugin store the dashboard reads. So a freshly enabled widget/hero
plugin (e.g. Koffi) only appeared after a full reload. refresh() now also reloads
the plugin store, so consumers pick the change up live.

* fix(trips): refresh the roster after a guest/member change

Adding an accountless guest (or any member change) only refreshed the members
modal's own list; the planner's tripMembers — which Costs participants, Collab,
etc. read from — was stale until a full reload. Expose refreshMembers from the
planner and let the members modal call it after each roster mutation.

* fix(security): hide co-members' private packing items in summary and share

getTripSummary built its packing section with listPackingItems(tripId) and no
viewer id, so the per-viewer #858 privacy filter never ran — the MCP
get_trip_summary tool then returned every other member's private/personal
packing items (name, category, owner) to any trip member or packing-scoped
token. Thread the acting user through so the filter applies. Separately, the
public trip share loaded packing with a raw SELECT * and published it when
share_packing was on; a public viewer is neither owner nor recipient, so only
Common (is_private = 0) items may surface.

* fix(plugins): forward webhook raw body as base64 and bound the ICS tz cache

The inbound webhook raw body was forwarded as a UTF-8 string, which corrupts a
non-UTF-8 signed payload (U+FFFD substitution) so a plugin can't reconstruct the
exact bytes for HMAC verification; forward base64 (rawBodyBase64) instead. Also
bound the ICS timezone-validity cache: its key is a free-form plugin/importer
zone string, so cap the distinct entries rather than growing for the process
lifetime.

* fix(atlas): resolve to a real neighbour polygon before a loose fallback box

A polygon-less micro-territory box (PS, XK) is far larger than the enclave it
represents and was winning the smallest-box tie-break over the sovereign whose
polygon actually contains the point, so cities inside a neighbouring country
resolved to the wrong territory. Only a tight box now auto-wins; a loose one
defers behind real polygons and is used only when no polygon contains the point.

* fix(feeds): include shared member trips in the all-trips calendar feed

buildUserIcs selected owned trips only, so a user's aggregate feed silently
dropped every trip shared with them as a member. Match the single-trip feed's
membership check.

* fix(plugins): tighten the OAuth pre-check and backup restore/snapshot

OAuth token-endpoint fast-fail now also rejects IPv6-literal loopback/metadata
hosts and .internal names (LAN stays reachable for a self-hosted IdP). Backup
restore enforces the decompressed-size cap by counting extracted bytes instead
of trusting the archive's declared sizes, and refuses a zip-slip path. A plugin
snapshot whose fold-in fails now keeps its -wal/-shm so no committed data is lost.

* fix(plugins): validate bag roster, reap orphaned erasures, harden brokers

A packing bag can only be assigned to a member of its trip. Queued user-erasures
for an uninstalled plugin are reaped instead of lingering forever. GDPR export
folds in host-side settings/oauth presence (best-effort). notify.send builds its
link by URL-parse so a backslash can't slip an off-site redirect through, and
dev-link discovery skips a Windows junction unless dev-link mode is on.

* fix(trips): refresh trip members only after a mutation, not on modal open

loadMembers notified the planner on every call, including the open effect, firing
a redundant refresh. Only mutation handlers notify now.

* docs(wiki): document the 3.3.0 plugin platform surface

Cover the plugin activity log, widget slots and tab takeover, the notify/ai/oauth
brokers, provider hooks, scheduler, GDPR hooks, the new plugin env vars, plugin
contributions on the map/atlas/PDF/journal pages, plugin runtime hardening, and
the Budget->Costs rename with multi-currency and CSV export.

---------

Co-authored-by: jubnl <jgunther021@gmail.com>
Co-authored-by: Dieter Blomme <dieterblomme@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: sld272 <zjrdmczh@outlook.com>
Co-authored-by: Nguyen Trong Binh <nguytb15@VN1N07HO1CD1015.local>
2026-07-11 19:02:34 +02:00

62 KiB
Raw Blame History

Plugin Development

Build a plugin with the trek-plugin-sdk package. A plugin is a directory with a manifest (trek-plugin.json), a built server entry, and — for page/widget plugins — a static client bundle. TREK runs your server code in an isolated child process and reaches it only over RPC; the browser part runs in a sandboxed, opaque-origin iframe. There is no other way in or out.

Scaffold

npx trek-plugin-sdk create                # interactive wizard
npx trek-plugin-sdk create my-plugin --type integration|page|widget|trip-page   # or direct
cd my-plugin

The wizard (run create with no name) asks for the id, type, author and permissions; the direct form takes them as flags.

This emits:

my-plugin/
  trek-plugin.json      # manifest
  package.json          # CommonJS marker + the SDK as a devDependency
  server/index.js       # your plugin code (built, plain JS)
  client/index.html     # native UI via the design kit (page / widget / trip-page only)
  README.md             # fill this in — the registry requires a screenshot

Run it locally with hot reload

npx trek-plugin-sdk dev        # http://localhost:4317

dev works straight after create — no npm install needed, because it injects require('trek-plugin-sdk') from the CLI itself, exactly like TREK injects it in production. It loads your server/index.js through the same definePlugin contract the host uses and gives you a real request loop without a full TREK: a dashboard listing your routes, the routes served under /api/<path>, your page/widget UI at /ui, a themed host preview at /preview (a real sandboxed frame with a theme/accent/appearance toggle, trek.invoke() proxied to your routes), and a reload on every save. The injected ctx is the full surface — every capability (costs, packing, files, notify, ai, settings, scheduler, meta, oauth, weather, rates, journal, db.tx, …), not just routes + db:own — and enforces exactly the permissions your manifest grants: an ungranted call throws PERMISSION_DENIED, so you catch a missing grant here rather than after install. db:own is backed by a real SQLite file (.trek-dev/db.sqlite) when the runtime has node:sqlite; every other capability is served from your fixtures with the same rules as production.

  • Hit a route as an unauthenticated request with ?_anon=1 (an auth: true route then returns 401, mirroring the host).
  • Feed the fixtures by dropping a dev-fixtures.json next to the manifest — it takes the same shape as createMockHost options, so you can seed trips, users, costs, settings, weather, canned AI results, etc.: { "actingUserId": 1, "trips": { "1": { "members": [1], "data": { … }, "costs": [ … ] } }, "users": {} }.
  • Fire a non-route entry point with GET /__dev/fire/<kind>[/<name>][/<fn>] (a JSON body or query params become the payload/args): /__dev/fire/job/refresh, /__dev/fire/scheduled/daily, /__dev/fire/event/place:created, /__dev/fire/hook/tripCardProvider/getCards, /__dev/fire/deleteUserData?userId=1 — so jobs, scheduled timers, event subscriptions and provider hooks are all testable in the same hot-reload loop as your routes.

dev above is fast but its host data is synthetic (fixtures / a scratch db:own). When you need your plugin to run against real trips, places, reservations, costs and real membership/permissions, link it into a running TREK instance instead of packing + uploading it every time.

On the server (a local trek-dev or a dev instance — never production), set:

TREK_PLUGINS_DEV_LINK=1

With the flag set, the Admin → Plugins tab shows a “Link a local plugin” field. Paste the absolute path to your built plugin directory (the one holding trek-plugin.json + server/index.js — build first, the loader runs the compiled artifact, not TS source) and hit Link. The plugin is symlinked into the plugins volume and registered INACTIVE, and shows a Dev-link badge in the list.

Prefer scripting? The same thing over HTTP:

curl -XPOST /api/admin/plugins/link -H 'content-type: application/json' \
  -d '{ "path": "/abs/path/to/your/plugin" }'

Activate it in the admin UI and consent to its permissions as usual. It now runs through the same capability RPC host as any installed plugin: real, membership-gated data, the acting user resolved host-side, no impersonation — code origin never touches the security gate.

Hot-reload: rebuild your plugin (e.g. tsc --watch emitting server/index.js) and TREK re-forks it automatically (a file-watch on the linked dir). To force it, POST /api/admin/plugins/:id/reload, or just hit Restart in the admin UI — both re-fork the child, picking up the new code (a rebuilt manifest that widened permissions still requires explicit re-consent).

Dev-link is gated behind TREK_PLUGINS_DEV_LINK on top of admin + the plugin kill-switch, and is off by default. It loads unsigned local code that mutates live between restarts, and under npm run dev the OS permission jail is off — so only ever enable it on a machine you control, pointed at a dev instance.

The plugin types

  • integration — background logic (jobs, routes) with no UI of its own. Most provider hooks are live (placeDetailProvider, warningProvider, tableContributor, mapMarkerProvider, pdfSectionProvider, atlasLayerProvider, journalEntryProvider); only photoProvider / calendarSource are declared in the SDK but not yet wired into the host — see Provider hooks.
  • page — adds a nav entry that opens a full-page sandboxed iframe.
  • widget — adds a card to the dashboard (sidebar slot), a hero-bar overlay (hero slot), a panel inside the trip planner's place-detail view (place-detail slot — the frame also receives the open placeId in trek:context, so it can show place-specific info like reviews or ratings), a panel inside the day-detail view (day-detail slot — receives the open dayId; the home for per-day content like outfit planning, live flight status or logistics), or a panel at the foot of a booking card in the reservation-detail view (reservation-detail slot — receives the open reservationId, for things like live check-in status or a seat map). Set the slot in capabilities.widget.slot.
  • trip-page — adds a tab inside every trip planner, so your UI lives in the trip alongside Plan / Transports / Files. The frame is the same sandboxed iframe as a page, but it receives the current tripId in trek:context (so you can scope data to the open trip) and it has no dashboard nav entry. The tab shows on desktop and mobile. Via capabilities.tripPage the plugin can also take over core tabs: replaces: ['transports', 'buchungen', …] hides the named tabs while the plugin is active (they return the moment it's deactivated; plan is never replaceable), and position picks the 0-based index of your tab in the bar instead of appending it. A plugin that replaces tabs gets a "Replaces planner tabs" chip in the admin list, so the takeover is visible before activation.

The SDK package

trek-plugin-sdk is injected at runtime — the host makes require('trek-plugin-sdk') resolve inside the child, so do not vendor it into your artifact. Add it as a devDependency only, so you get types, createMockHost for tests, and the trek-plugin CLI:

npm i -D trek-plugin-sdk

Writing the server

Your server/index.js exports a definePlugin(...) object. Everything reaches TREK through the ctx argument.

const { definePlugin } = require('trek-plugin-sdk')

module.exports = definePlugin({
  // Runs once when the plugin is activated. NOTE: onLoad has no user context —
  // ctx.trips.* is refused here (see the ctx table).
  async onLoad(ctx) {
    await ctx.db.migrate('001_init', 'CREATE TABLE IF NOT EXISTS cache (k TEXT PRIMARY KEY, v TEXT)')
    ctx.log.info('loaded')
  },

  // Runs once on deactivation/stop. Use it to flush or release resources.
  async onUnload(ctx) {
    ctx.log.info('unloading')
  },

  // HTTP routes, mounted at /api/plugins/<id><path>.
  routes: [
    { method: 'GET', path: '/status', auth: true, async handler(req, ctx) {
      const rows = await ctx.db.query('SELECT COUNT(*) AS n FROM cache')
      return {
        status: 200,
        headers: { 'content-type': 'application/json' },
        body: JSON.stringify({ n: rows[0].n, user: req.user?.username }),
      }
    }},
  ],

  // Scheduled jobs — TREK owns the cron and calls your handler on the schedule.
  // Requires the `jobs:run` permission (opt-in: scheduled work runs with NO user,
  // so its trip reads are refused — a job can only use ctx.db and declared egress).
  // An invalid cron expression is skipped; jobs stop when the plugin is deactivated.
  jobs: [
    { id: 'refresh', schedule: '*/15 * * * *', async handler(ctx) { /* … */ } },
  ],

  // Persistent, userless timers you arm at runtime with ctx.scheduler — they
  // survive restarts and call this handler when due. Also needs `jobs:run`.
  //   await ctx.scheduler.in(3_600_000, 'remind', { tripId })   // once, in 1h
  //   await ctx.scheduler.every(86_400_000, 'digest')           // daily
  async scheduled({ name, payload }, ctx) {
    if (name === 'remind') { /* … one-shot fired; re-arm if you want another */ }
  },

  // GDPR data-subject rights (needs `hook:user-data`). Userless — you only get the
  // userId and act on your OWN db. deleteUserData is delivered DURABLY (queued and
  // retried until it succeeds, even across restarts), so make it idempotent.
  async deleteUserData({ userId }, ctx) {
    await ctx.db.exec('DELETE FROM my_prefs WHERE user_id = ?', userId)
  },
  async exportUserData({ userId }, ctx) {
    return await ctx.db.query('SELECT * FROM my_prefs WHERE user_id = ?', userId)
  },
})

The routes and job ids you declare here are the authoritative ones: the host reads them off your loaded definition (a route's array index is its internal id). The routes block the scaffold writes into trek-plugin.json is only a declaration for readers — the manifest parser does not consume it.

The ctx object

Area Methods Requires
ctx.db query(sql, …args) / exec(sql, …args) / migrate(id, sql) / tx(ops) against your own SQLite file. tx([{sql, args?}, …]) runs up to 100 statements in one transaction (all commit or all roll back; reads see the batch's own earlier writes) → `{ results: [{changes?} {rows?}, …] }`
ctx.trips getById / getPlaces / getReservations / getDays / getAccommodations / listMine() — enumerate every trip the acting user can access (membership-checked). getDays includes each day's assignments + notes_items; getReservations includes endpoints + day_positions db:read:trips
ctx.trips.update(tripId, fields) update trip fields (title/dates/currency/reminder_days/…) db:write:trips
ctx.trips.create(input) create a new trip owned by the acting user (importers) — title required, plus description?/start_date?/end_date?/currency?/reminder_days?/day_count? db:create:trips (+ trip_create)
ctx.places create(tripId, fields) / update(tripId, placeId, fields) / delete(tripId, placeId) db:write:places
ctx.days create(tripId, {date?, notes?}) / update(tripId, dayId, {notes?, title?}) / delete(tripId, dayId) db:write:days
ctx.itinerary assign(tripId, dayId, placeId, notes?) / unassign(tripId, assignmentId) — place↔day db:write:itinerary
ctx.meta get / set / list / delete your own namespaced data on a trip/place/day (enrich core entities without forking the schema) db:meta
ctx.packing list(tripId) — a trip's packing items (membership-checked, respects private-item visibility) db:read:packing
ctx.files list(tripId) — a trip's files, trash excluded (membership-checked) db:read:files
ctx.files.getContent getContent(tripId, fileId) → { name, mimetype, size, content_base64 } — a file's bytes as base64 (10MB cap; trashed files refused) db:read:files:content
ctx.journal listMine() — the acting user's own travel journals; getEntries(journeyId) — one journey's entries (photos/story/checkins, access-checked) db:read:journal (+ Journey addon)
ctx.atlas visited() — the acting user's visited countries + regions; bucketList() — their bucket-list items db:read:atlas (+ Atlas addon)
ctx.vacay mine() — the acting user's vacation plan db:read:vacay (+ Vacay addon)
ctx.collections listMine() / get(id) — the acting user's saved-place collections db:read:collections (+ Collections addon)
ctx.collections (write) create(input) / update(id, input) / savePlace(input) / copyToTrip(input) / deletePlace(placeId) — per-collection role enforced by the service db:write:collections (+ Collections addon)
ctx.atlas (write) markCountry(code) / unmarkCountry / markRegion(regionCode, countryCode, regionName?) / unmarkRegion / createBucketItem(input) / deleteBucketItem(itemId) — the acting user's own rows db:write:atlas (+ Atlas addon)
ctx.vacay (write) toggleEntry(date) — the acting user's own PTO day on their active plan; toggleCompanyHoliday(date, note?) db:write:vacay (+ Vacay addon)
ctx.journal (write) createEntry(journeyId, {entry_date, ...}) / updateEntry(entryId, fields) / deleteEntry(entryId) — owner/contributor-gated; createJourney({title, subtitle?, trip_ids?}) / deleteJourney(journeyId) — a new journal owned by the acting user (importers bootstrap the journal they then fill) db:write:journal (+ Journey addon)
ctx.files (write) create(tripId, {name, content_base64, mimetype?, ...}) (10MB cap, blocked extensions refused) / createLink(tripId, fileId, opts) / update / softDelete — broadcasts file:* db:write:files (+ the acting user's file_upload/file_edit/file_delete)
ctx.collab listNotes(tripId) / listPolls(tripId) / listMessages(tripId, before?) — a trip's notes, polls (with options + voters) and chat (newest 100, oldest first; before = a message id to page back), membership-checked db:read:collab (+ Collab addon)
ctx.collab (write) createNote(tripId, {title, ...}) / createPoll(tripId, {question, options}) / votePoll(tripId, pollId, optionIndex) / createMessage(tripId, text, replyTo?) — broadcasts collab:* db:write:collab (+ collab_edit, Collab addon)
ctx.trips.addMember / .removeMember addMember(tripId, userId) — grants trip access; removeMember(tripId, userId) — revokes it (never the owner), so a directory-sync integration can reconcile departures too db:write:members (+ member_manage)
ctx.notify send({title, body, link?, scope, targetId}) — bell inbox + email/ntfy fan-out; recipient forced to the acting user (scope:'user') or a trip they belong to (scope:'trip') notify:send
ctx.ai complete(prompt, system?) → { text }; extract(text, jsonSchema, prompt?) → { results } — the admin/user-configured provider; host holds the key; output is DATA (no auto-writes) ai:invoke
ctx.oauth getAccessToken() → a short-lived access token for the acting user of a third-party service the host connected on their behalf (Settings → Plugins → Connect); null if not connected / userless. Host holds the refresh token + client secret oauth:client
ctx.scheduler at(whenMs, name, payload?) / in(ms, name, payload?) / every(ms, name, payload?) / cancel(name) — persistent, userless timers that survive restarts and fire your scheduled(input, ctx) handler. set is an upsert by name; caps: ≤100 tasks, 8 KB payload, recurring interval ≥ 60 s, ≤ ~1 year out. Same risk class as jobs (no acting user → trip reads refused) jobs:run
ctx.settings get(key) — the acting user's own value for one of your scope:'user' settings fields (decrypted host-side). Returns undefined for an unset value or a userless context (job/onLoad) — fall back to ctx.config there. Users fill these in under Settings → Plugins; secrets are stored encrypted and never echoed back none (your own settings)
ctx.daynotes list(tripId, dayId) — a day's notes (membership-checked) db:read:daynotes
ctx.daynotes (write) create(tripId, dayId, {text, time?, icon?, sort_order?}) / update(tripId, dayId, noteId, fields) / delete(tripId, dayId, noteId) — broadcasts dayNote:* db:write:daynotes
ctx.packing (write) create(tripId, {name, category?, checked?, is_private?, visibility?, recipient_ids?}) / update(tripId, itemId, fields) / delete(tripId, itemId) — broadcasts packing:*, private items (#858) stay owner-scoped db:write:packing
ctx.packing (bags) listBags(tripId) / createBag(tripId, {name, color?}) / updateBag / deleteBag / setBagMembers(tripId, bagId, userIds) — bags carry no privacy db:write:packing
ctx.weather get(lat, lng, date?) — the host's cached forecast (tenant-free) weather:read
ctx.rates get(base) → a map of quote → rate relative to base (e.g. 'EUR' → { USD: 1.08, … }; cached upstream, tenant-free); null on an upstream failure rates:read
ctx.categories list() — the global place-category list db:read:categories
ctx.tags list() / create({name, color?}) / update(tagId, fields) / delete(tagId) — the acting user's own tags db:read:tags / db:write:tags
ctx.trips.members members(tripId) — the trip roster (id + display fields), membership-checked db:read:trips
ctx.todos list(tripId) / create(tripId, {name, ...}) / update(tripId, todoId, fields) / delete(tripId, todoId) — broadcasts todo:* db:read:todos / db:write:todos (+ packing_edit)
ctx.costs getByTrip(tripId) / listMine() — read budget items (membership-checked) db:read:costs
ctx.costs (write) create(tripId, input) / update(tripId, itemId, input) / delete(tripId, itemId) — broadcasts budget:* db:write:costs
ctx.reservations listMine() — every booking across the acting user's accessible trips (membership-checked) db:read:trips
ctx.reservations (write) create(tripId, input) / update(tripId, reservationId, input) / delete(tripId, reservationId) — full parity with the app (accommodation, budget-sync, notification, broadcasts reservation:*). input.endpoints = array of { role: 'from'|'to'|'stop', name, lat, lng, code?, sequence?, timezone?, local_time?, local_date? }; update semantics: omitted = keep, [] = delete all, array = replace (endpoint ids are not stable) db:write:reservations
ctx.accommodations create(tripId, { place_id, start_day_id, end_day_id, check_in?, check_in_end?, check_out?, confirmation?, notes? }) / update(tripId, accommodationId, fields) / delete(tripId, accommodationId) — lodging blocks; create auto-creates the partner hotel reservation, delete cascades it (broadcasts like the app) db:write:accommodations (+ day_edit)
ctx.users getById(id) — public profile only (id, username, display_name, avatar) db:read:users
ctx.ws.broadcastToTrip(tripId, event, data) broadcast to a trip's members (event forced to plugin:<id>:<event>) ws:broadcast:trip
ctx.ws.broadcastToUser(userId, event, data) broadcast to one user ws:broadcast:user
ctx.plugins.call(id, fn, args?) call a function another plugin exposes and get its result — id must be a declared, satisfied pluginDependency that lists fn in its capabilities.provides a plugin dependency (no permission)
ctx.events.emit(name, payload?) publish an event to dependents that subscribed — name must be in your capabilities.emits — (no permission)
ctx.config your resolved settings (secrets delivered decrypted) —
ctx.log info / warn / error → your error log —
ctx.id your plugin id (string) —

Calling a method your manifest didn't grant returns PERMISSION_DENIED; a method the host doesn't expose at all returns UNKNOWN_METHOD.

ctx.trips only works inside a route handler. The host binds the acting user from the authenticated request and membership-checks every trip read against it. onLoad and jobs have no user, so their trip reads are refused with RESOURCE_FORBIDDEN. The SDK's getById(tripId, asUserId?) signature keeps an asUserId parameter for source compatibility, but the host ignores it — you cannot read another user's trips by passing an id.

Writes (ctx.trips.update / ctx.places / ctx.days / ctx.itinerary / ctx.costs.create) are route-context only too, and doubly gated: the host checks the acting user can access the trip AND holds the app's edit permission for that entity (place_edit / day_edit / trip_edit), exactly like the web UI. They run through the same services and broadcast the same events, so open sessions update live. Input is validated against TREK's own schemas (a bad payload is BAD_PARAMS), and every write is recorded in the tamper-evident capability audit log against the acting user. A plugin can only change what its user could change by hand.

ctx.costs ("costs" = budget items) behaves exactly like ctx.trips: reads are membership-checked against the request's user and only work inside a route handler (onLoad/jobs have no user → RESOURCE_FORBIDDEN). getByTrip(tripId) returns one trip's budget items (hydrated with members/payers); listMine() aggregates budget items across every trip the acting user can access. create/update/delete(tripId, …) mutate a trip's budget items — gated exactly like a normal budget write (the same model the planner write scopes db:write:places/days/itinerary/trips use): the acting user needs the budget_edit permission on that trip, the input is validated against TREK's budget schema, and a successful create broadcasts the same budget:created event the app emits. Every ctx.costs.* call also requires the Costs (budget) addon to be enabled — if the admin has turned it off, the call is refused with RESOURCE_FORBIDDEN.

Route auth

Routes are authenticated by default (req.user is the logged-in user). Set auth: false for OAuth callbacks or webhooks that can't carry a session.

The proxy forwards { method, path, query, body, headers, user }. req.headers is populated ONLY on auth: false routes (an authenticated route gets {}) and only an explicit, credential-free allowlist — the common provider signature + event headers (stripe-signature, x-hub-signature-256, svix-signature, x-gitlab-event, content-type, user-agent, …). Cookie, Authorization, X-Socket-Id and every session/forwarded-auth header are stripped and never reach your code, so a forwarded header can't leak a TREK session. To trust a webhook, verify the provider's signature over the raw body against a secret you hold in ctx.config (admin-set instance setting) or ctx.settings (per-user).

Writing the client (page / widget)

The iframe is served same-origin from /plugin-frame/<id>/… but sandboxed without allow-same-origin, so it runs at an opaque origin: it can't read cookies or the parent DOM. It talks to TREK only via postMessage (target origin must be '*' — an opaque frame has no nameable origin).

Your whole client/ directory is served (and shipped by pack) — not just index.html. The frame CSP allows the plugin its own static files (and nothing from any other host), so a multi-file build works as-is: point a bundler at client/ and use relative asset paths (Vite: base: './'), drop the output in, done. That includes React/Vue/Svelte builds — no more inlining the bundle into index.html (inline still works, and is what the design-kit marker does).

Because the frame can't load TREK's stylesheet, we ship it. Drop one line in your client/index.html <head>:

<!-- trek:ui -->

dev and pack expand that marker into the inlined TREK design kit — a token-driven stylesheet plus a window.trek bridge. It costs nothing to keep the source a one-liner, and a rebuild always ships the current kit. The kit:

  • gives you native components — glass panels, cards, buttons, inputs, chips, list rows, hover — that swap correctly between light and dark;
  • follows the user's live accent scheme, custom accent and high-contrast (it applies the tokens TREK sends);
  • mirrors the host's appearance flags (reduced-motion, no-transparency, density);
  • auto-reports your height (widgets/pages self-size — no manual trek:resize);
  • installs window.trek so you never hand-roll postMessage.

window.trek also carries trek.ui — tiny DOM builders that emit kit-styled elements, so you can build UI with no bundler and no CSS:

const { ui } = trek
ui.mount(ui.card([
  ui.el('div', { class: 'trek-title', text: 'Nearby' }),
  ui.button('Refresh', { variant: 'primary', onClick: refresh }),
  ui.chip('open now', 'success'),
]))
// ui.el(tag, props, children) is the general builder; props take class/text/html/on:{event}.

The scaffold seeds a working example. A minimal client:

<!doctype html>
<html lang="en">
<head>
  <meta charset="utf-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1" />
  <!-- trek:ui -->
</head>
<body>
  <div class="trek-glass trek-stack" style="margin:16px">
    <div class="trek-title">Your plugin</div>
    <p class="trek-muted" id="hello">…</p>
    <button class="trek-btn trek-btn--primary" id="go">Say hello</button>
  </div>
  <script>
    trek.onContext((ctx) => { document.getElementById('hello').textContent = 'theme: ' + ctx.theme })
    document.getElementById('go').addEventListener('click', async () => {
      try { const data = await trek.invoke('/hello'); document.getElementById('hello').textContent = 'Hello ' + data.hello }
      catch (e) { trek.notify('error', e.message) }
    })
  </script>
</body>
</html>

Component classes (the bootstrap adds trek-ui to <body>):

Class What
.trek-glass the signature frosted-glass surface
.trek-card a solid card
.trek-interactive add to a glass/card for the native hover-lift
.trek-btn + --primary / --secondary / --ghost / --danger buttons
.trek-input / .trek-textarea / .trek-select / .trek-label form controls
.trek-chip + --accent / --success / --danger / --warning / --info chips / badges
.trek-row a hover-highlight list row
.trek-title / .trek-muted / .trek-faint text helpers
.trek-stack / .trek-cluster vertical / horizontal flex with gap
.trek-menu-enter / .trek-menu-enter-left dropdown enter (scales from its trigger corner)
.trek-popover-enter / .trek-modal-enter / .trek-backdrop-enter / .trek-toast-enter the host's enter animations (trek-modal-enter becomes a bottom drawer below 640px)
.trek-page-enter subtle fade-up on mount
.trek-stagger children fade up with a 40ms cascade
.trek-skeleton shimmer loading placeholder
.trek-pie-reveal / .trek-bar-fill chart reveal animations (plus the trek-progress-fill keyframe, driven by --trek-progress-to)

All animations honour reduced motion: under [data-reduce-motion] or prefers-reduced-motion they degrade to a gentle 120ms fade, and the skeleton stops shimmering.

Selects are auto-upgraded. With the kit inlined, every native <select> becomes a host-styled, keyboard-accessible dropdown that matches TREK — the OS-drawn popup can't be themed, so the kit replaces it while keeping the real <select> as the value/form source (it still fires change). Write a plain <select> and it just works. Add data-trek-native to a field to keep the browser default; multiple and size selects are always left native.

The window.trek bridge:

Call Does
trek.onContext(cb) run cb(context) now (if already received) and on every update; returns an unsubscribe fn
trek.context the last context (or null)
trek.invoke(sub, { method, body }) call your own route; returns a Promise (rejects with an Error, .code = HTTP status)
trek.notify(level, message, duration?) toast (info/success/warning/error); optional duration in ms (clamped 1.5–15s)
trek.confirm({ title?, message, confirmLabel?, cancelLabel?, danger? }) host-rendered native confirm dialog; resolves true/false (one at a time — a second concurrent request resolves false)
trek.navigate(to) in-app navigation (relative paths only)
trek.openExternal(url) open an http(s) URL in a new tab (the sandbox itself can't)
trek.onEvent(cb) cb(event, tripId) for core events on the trip in view — names only, no payloads; refetch via invoke(); returns an unsubscribe fn
trek.resize(px) override the auto height (ignored on full pages — see trek:resize below)
trek.ready() / trek.requestContext() re-handshake / re-request the context

Preview it: npx trek-plugin-sdk dev, then open /preview — it renders your UI in a real sandboxed frame with a theme/accent/appearance toggle and proxies trek.invoke() to your routes.

The raw bridge (without the kit)

If you'd rather not use the kit, talk to the frame yourself. Announce readiness and handle messages:

window.parent.postMessage({ type: 'trek:ready' }, '*') // TREK replies with trek:context
window.addEventListener('message', (e) => {
  if (e.source !== window.parent) return          // opaque frame: trust the parent window
  const m = e.data
  if (m.type === 'trek:context') { /* m.theme, m.tokens, m.appearance, … (below) */ }
  if (m.type === 'trek:response') { /* m.requestId, m.data */ }
  if (m.type === 'trek:error')    { /* m.requestId, m.code, m.message */ }
})
window.parent.postMessage({ type: 'trek:invoke', requestId: '1', sub: '/status', method: 'GET' }, '*')

Messages you send to TREK:

Message Payload Effect
trek:ready — TREK replies with trek:context
trek:context:request — re-request the context
trek:navigate { to } in-app navigation (relative paths only)
trek:notify { level, message, duration? } toast; level = info/success/warning/error; duration in ms, clamped 1500–15000
trek:resize { height } set the iframe height (capped at 2000px); ignored for full pages (page / trip-page), which always fill their host container
trek:invoke { requestId, sub, method, body } call your own route; resolves as trek:response or trek:error
trek:confirm { requestId, title?, message?, confirmLabel?, cancelLabel?, danger? } host-rendered ConfirmDialog; answered as trek:confirm:result
trek:openExternal { url } open an http(s) URL in a new noopener tab; anything else is dropped

Messages TREK sends you:

Message Payload
trek:context { tripId, placeId, dayId, reservationId, userId, theme, locale, dir, hostOrigin, user, formats, tokens, appearance } (see below) — re-sent whenever the theme, appearance, locale or formats change
trek:response { requestId, data } — a successful trek:invoke
trek:error { requestId, code, message } — a failed trek:invoke (code is the HTTP status or "error")
trek:confirm:result { requestId, confirmed } — the user's answer to your trek:confirm
trek:event { event, tripId } — a core event fired on the trip in view; names only, never payloads — refetch what you need via trek:invoke

The frame's CSP is locked down per plugin: default-src 'none', own inline scripts/styles + the plugin's own /plugin-frame/<id>/ files only (no other host may serve it script/style/img), connect-src limited to the hosts you were granted via http:outbound:<host> permissions (not merely the egress[] you declared), no popups.

The context payload

Field Type
tripId string | null — the trip in view (a trip-page tab, or a widget on a trip), else null. IDs arrive as strings — compare with String(id)
placeId string | null — the place in view (a place-detail slot), else null
dayId string | null — the day in view (a day-detail slot), else null
reservationId string | null — the reservation in view (a reservation-detail slot), else null
dir 'ltr' | 'rtl' — the host's text direction; the kit mirrors it onto your <html> so RTL hosts get RTL plugin UIs
userId string | null
theme 'light' | 'dark'
locale e.g. 'en'
hostOrigin the app origin
user { name, avatar, isAdmin } | null — never an email; role only as a boolean
formats { locale, currency, timeFormat, distanceUnit, temperatureUnit, timezone }
tokens TREK's resolved CSS design tokens for the current theme (see below)
appearance { scheme, density: 'comfortable'|'compact', reducedMotion, noTransparency }

Matching the TREK look by hand (m.tokens)

tokens is the whole global palette resolved for the current theme — surfaces (--bg-card, --bg-hover, …), text (--text-primary/-secondary/-muted/-faint), borders, the accent family (--accent, --accent-text, --accent-hover, --accent-subtle), semantic + soft fills (--success/--danger/--warning/--info -soft), shadows (--shadow-*), radii (--radius-*) and fonts (--font-system). Apply them as CSS variables and your UI matches the host exactly — in both themes and under a custom accent or high-contrast — instead of hard-coding a palette that drifts:

function applyContext(m) {
  document.documentElement.dataset.theme = m.theme                 // for your dark rules
  for (const k in m.tokens) document.documentElement.style.setProperty(k, m.tokens[k])
  const a = m.appearance || {}
  document.documentElement.toggleAttribute('data-reduce-motion', !!a.reducedMotion)
  document.documentElement.toggleAttribute('data-no-transparency', !!a.noTransparency)
}
// in your trek:context handler: applyContext(m)

tokens/appearance are non-secret display values only, re-sent on every theme or appearance change so plugins feel native rather than bolted-on. (The glassy tokens the dashboard uses — --glass-*, --r-*, --sh-* — aren't in tokens; the design kit bakes those, since they only change with light/dark, not the accent.) Honour appearance.reducedMotion / noTransparency, and the frame also inherits the OS prefers-reduced-motion. Dashboard widgets are wrapped in the native glassy tool card and auto-size to the height you report via trek:resize, so render flush and transparent — the design kit reports your height for you.

Settings

Declare settings in the manifest; TREK renders the form (you write no settings UI). scope: "instance" settings are set once by the admin; scope: "user" settings are per-user. secret: true fields are stored encrypted and delivered decrypted through ctx.config (server-side only) — never to the iframe. Resolved values arrive in ctx.config.

Host-brokered OAuth (ctx.oauth)

A plugin can act as an OAuth client of a third-party service without ever handling the secrets. The host runs the entire flow — authorize → callback → token exchange → refresh — with PKCE (S256) and a state check, and holds the tokens. The plugin only triggers "connect" and reads a short-lived access token at runtime.

Setup. Declare the provider as scope: "instance" settings the admin fills in: oauth_authorize_url, oauth_token_url, oauth_scopes (optional), plus the two secret: true fields oauth_client_id and oauth_client_secret. (A settings field may also carry an oauth: { initPath, callbackPath } block.)

Connecting. A user connects under Settings → Plugins → Connect, which sends them to the provider's authorize page. The callback returns to …/api/plugin-oauth/<id>/callback; the host verifies the state (single-use, 10-minute TTL, bound to that user — CSRF defence), exchanges the code and stores the tokens per user, encrypted at rest.

Using it. Read the access token in a route handler:

const token = await ctx.oauth.getAccessToken() // needs the `oauth:client` permission
if (!token) return { status: 401, body: 'connect this plugin first' }
// call the third-party API with `Authorization: Bearer ${token}`
  • Returns null when the acting user hasn't connected, or in a userless context (a job / onLoad).
  • The host auto-refreshes a token that is expiring (60-second skew) using the refresh token it holds.
  • The plugin never sees the refresh token or the client secret.
  • The authorize/token URLs must be https to a non-private host, and the host-side token exchange goes through the SSRF guard (blocks the link-local / cloud-metadata range and pins the resolved IP against DNS-rebind; a self-hosted internal IdP on loopback/LAN still works).

GDPR data-subject hooks

Grant hook:user-data and implement either handler to honour data-subject rights. Both are userless — the plugin only receives the userId and acts on its own ctx.db:

module.exports = definePlugin({
  permissions: ['db:own', 'hook:user-data'],
  async deleteUserData({ userId }, ctx) {           // erasure
    await ctx.db.exec('DELETE FROM my_prefs WHERE user_id = ?', userId)
  },
  async exportUserData({ userId }, ctx) {           // portability / access
    return await ctx.db.query('SELECT * FROM my_prefs WHERE user_id = ?', userId)
  },
})
  • Erasure is durable. When a user is deleted, deleteUserData is queued and retried until the plugin ACKs, across restarts and even after the plugin is later reactivated — so make it idempotent.
  • Export is complete-or-flagged. A data-access request aggregates exportUserData across every granted plugin. A plugin that is currently inactive but holds hook:user-data, or one whose export errored/timed out, is flagged pending (never silently omitted), so the admin knows to reactivate it to finish the request.
  • Uninstalling a plugin also purges its stored OAuth tokens and state.

Provider hooks

A hook is core calling into your plugin for data (host→plugin). Declare it on the plugin definition and grant the matching hook:* permission:

module.exports = definePlugin({
  hooks: {
    placeDetailProvider: {
      // Return extra rows TREK renders natively on a place. Runs with the current
      // user bound, on a short timeout — a slow/failing call is skipped, never fatal.
      async getDetails(placeId, ctx) {
        return [{ label: 'Crowd', value: 'Quiet now' }, { label: 'Guide', url: 'https://…' }]
      },
    },
  },
})
Hook Permission Status
placeDetailProvider.getDetails(placeId, ctx) → { label, value?, url? }[] hook:place-detail-provider live — shown in the place-detail panel; also GET /api/place-details/:placeId
warningProvider.getWarnings(tripId, ctx) → { level, message, dayId?, placeId? }[] hook:trip-warning-provider live — validation warnings shown as a non-blocking banner in the trip planner; also GET /api/trip-warnings/:tripId
tableContributor.getContributions(view, tripId, ctx) → TableContribution[] hook:table-contributor live — host-rendered columns/actions keyed by entityId in the reservations, transports, places, day, costs, packing, files and todos views. A column is {kind:'column', entityId, id, label, value?, url?, icon?, tone?} (url is http/https/mailto only); an action is {kind:'action', entityId, id, label, icon?, target} where target opens your sandboxed frame ({kind:'frame', sub}) or calls a route ({kind:'route', method, sub}). All fields are bounded + normalized host-side; also GET /api/view-contributions/:view/:tripId
mapMarkerProvider.getMarkers(tripId, ctx) → MapMarkerContribution[] hook:map-marker-provider live — bounded markers overlaid on the trip map (#587). Each is {id, lat, lng, label?, popupText?, url?, icon?, tone?}; coordinates are range-checked (−90..90 / −180..180), text length-capped, url http/https/mailto-only, count capped (≤200/plugin). Declarative only — plugin JS never runs on the map canvas. Also GET /api/map-markers/:tripId
pdfSectionProvider.getSections(tripId, ctx) → PdfSection[] hook:pdf-section-provider live — text-only sections appended to the trip PDF export. Each is {title, paragraphs?, table?}; the host escapes and lays everything out itself (no markup ever reaches the document), caps counts (≤5 sections/plugin, ≤20 paragraphs, ≤8 headers, ≤50 rows) + lengths (title 120, paragraph 2000, header 60, cell 200) and clips rows to the header width. Also GET /api/pdf-sections/:tripId
atlasLayerProvider.getLayers(ctx) → AtlasLayer[] hook:atlas-layer-provider live — country tint layers drawn over the Atlas world map (wishlists, advisories, …). User-scoped: the host binds the acting user, the hook takes no target parameter. Each layer is {id, name?, countries: [{code, tone?, label?}]}; codes must be ISO-3166 alpha-2 (uppercase-coerced), tone is enum-whitelisted, counts capped (≤3 layers/plugin, ≤300 countries/layer). Declarative only — plugin JS never runs on the map canvas. Also GET /api/atlas-layers
journalEntryProvider.getRows(entryId, ctx) → { label, value?, url? }[] hook:journal-entry-provider live — extra rows rendered under a journal entry card (needs the Journey addon; the entry's journey is access-checked like the journal detail routes). Same hardening as place details plus server-side normalization: ≤12 rows/plugin, label ≤60, value ≤200, url http/https/mailto-only. Also GET /api/journal-entry-rows/:entryId
tripCardProvider.getCards(tripIds, ctx) → TripCardContribution[] hook:trip-card-provider live — small badges on the dashboard trip cards. Called ONCE with all visible tripIds (each already access-checked for the acting user), returns { tripId, id, label, value?, icon?, tone?, url? }[]; the host bounds every field (label 64, value 256, tone enum, url http/https/mailto-only), caps the count (≤40/plugin) and drops any badge whose tripId wasn't requested. Declarative only. Also GET /api/trip-card-contributions?tripIds=…
photoProvider.search(query, {page, limit}, ctx) / .getById(id, ctx) hook:photo-provider live — plugin photo sources aggregated at GET /api/plugin-photos/search (+ /sources, /item) for the picker. Each {id, title?, thumbnailUrl, fullUrl, takenAt?}; thumbnail/full URLs must be http/https, per-source count capped, a failing source skipped
calendarSource.getName(ctx) / .getEvents(userId, start, end, ctx) hook:calendar-source live — plugin calendar events aggregated for the signed-in user at GET /api/plugin-calendar?start=&end=. Each {id, title, start, end, allDay} (ISO dates); count capped, a failing source skipped

Each hook method receives its args plus the per-invocation ctx, so any ctx.trips.* read it makes is membership-checked against the current user (like a route handler).

Event subscriptions

React to core activity with events + the events:subscribe permission. Handlers fire without a user (like a job) and receive { event, tripId, entity?, entityId?, snapshot? }: entity is the event family ('reservation', 'place', …), entityId says which entity changed, and snapshot is a whitelisted field view of the changed entity — delivered only when your plugin also holds the family's read grant (db:read:trips for place/day/reservation/accommodation/assignment/trip, db:read:costs for budget, db:read:packing for packing, db:read:daynotes for dayNote, db:read:files for file). Without that grant you get exactly the id hint, as before:

module.exports = definePlugin({
  permissions: ['events:subscribe', 'db:read:trips'],
  events: [
    { on: 'reservation:created', async handler({ event, tripId, entityId, snapshot }, ctx) {
        // with db:read:trips the snapshot carries the reservation's fields
        // (title, times, type, endpoints, ...) — no refetch needed
        await ctx.db.exec('INSERT INTO seen (trip, res, title) VALUES (?, ?, ?)', tripId, entityId, snapshot?.title ?? '')
    } },
    { on: '*', handler(e) { /* firehose: every core event */ } },
  ],
})

entityId is absent for bulk/reorder events and for events with no single entity; snapshot is additionally absent for deletes (nothing left to show) and for a private packing item (its broadcast is owner-scoped — #858). Never assume either is set. The snapshot whitelist never carries user ids (owner/paid-by/participants), trips.feed_token or other secrets — the id is only ever the changed entity's own id (never a parent or a user id).

Delivery is fire-and-forget on a short timeout, so a slow subscriber never blocks a core write. Because there's no user, trip reads (ctx.trips.*) are refused inside a handler — beyond the snapshot, use the plugin's own ctx.db, ctx.ws.*, or an outbound call. A plugin's own plugin:* broadcasts are never delivered back, so handlers can't loop. Common events: place:*, day:*, assignment:*, budget:*, file:*, accommodation:*.

Dependencies

A plugin can declare that it needs certain addons enabled, or other plugins installed, before it will run. Both are top-level manifest arrays, and both are enforced at activation — installing always succeeds, so a missing dependency is a fixable state, never a broken download.

requiredAddons

"requiredAddons": ["budget", "journey"]

Addon ids (see Addons Overview) that must be enabled for the plugin to activate. If one is off, enabling the plugin is refused and the admin panel names the addon to turn on. Turning a required addon off while the plugin is running auto-disables the plugin (and anything that depends on it) — a plugin never runs against a disabled addon. Ids are validated for shape only, so a plugin may name an addon a given TREK build doesn't have; it just stays un-activatable there.

pluginDependencies

"pluginDependencies": [
  { "id": "koffi", "version": ">=1.2.0 <2.0.0" }
]

Other plugins that must be installed and version-satisfied (a standard semver range) before this one activates. That range is the real contract for anything you call on the dependency (see Talking to other plugins).

Enforcement, all at activation time:

  • Missing dependency → activation is blocked and the panel offers a one-click download that fetches the newest registry version satisfying your range (pulling its own dependencies too), then retries.
  • Installed but out of range → same block; the panel offers to update it.
  • Installed but disabled → enabling your plugin auto-enables the dependency first, transitively (deepest dependency first).
  • Disabling a dependency cascades: every plugin that (transitively) depends on it is disabled too.
  • A dependency cycle (A → B → A) is refused with a clear error.

Dependencies are also resolved deps-first at boot, so a plugin's dependencies are already up before it starts.

Talking to other plugins

Isolation is the default — plugins can't see each other. To let a plugin be used by the plugins that depend on it, it opts in by declaring a surface in its manifest capabilities, and the host routes calls/events between the two child processes. There is no permission for this: authorization is the dependency edge itself — plugin A may call or subscribe to plugin B only if A declares B as a satisfied pluginDependency, and only for the names B publicly declares.

Exports — request / response

The dependency (B) exposes named functions and lists them in capabilities.provides:

// plugin "koffi"
module.exports = definePlugin({
  exports: {
    // `args` is whatever the caller passed; `ctx` is a per-call context.
    async convert({ amount, from, to }, ctx) {
      return { amount: amount * rate(from, to), to }
    },
  },
})
// manifest: "capabilities": { "provides": ["convert"] }

The dependent (A) declares koffi as a dependency and calls it:

// manifest: "pluginDependencies": [{ "id": "koffi", "version": ">=1.0.0 <2.0.0" }]
const out = await ctx.plugins.call('koffi', 'convert', { amount: 10, from: 'USD', to: 'EUR' })
  • A call is refused (RESOURCE_FORBIDDEN) if the target isn't a satisfied dependency, isn't currently active, or the function isn't in the target's provides.
  • The acting user is propagated: B's export runs as A's current user, so any ctx.trips.* read B makes is membership-checked against that user — B can't be tricked into reading data the calling user couldn't see.
  • The call is bounded by a timeout and recorded in the capability audit log (plugin:<target>#<fn>), attributed to A and the acting user.
  • B owns its contract: only functions in provides are reachable — routes, jobs and helpers stay private. Because your pluginDependencies range pins B's version, B can refactor internals freely and only breaks you on a major bump.

Events — publish / subscribe

The emitter (B) declares event names in capabilities.emits and publishes them:

// manifest: "capabilities": { "emits": ["rate.updated"] }
ctx.events.emit('rate.updated', { pair: 'USD/EUR', rate: 0.92 })   // fire-and-forget

A dependent (A) subscribes by naming the source plugin + event:

module.exports = definePlugin({
  subscriptions: [
    { plugin: 'koffi', event: 'rate.updated', async handler(payload, ctx) {
        await ctx.db.exec('UPDATE cache SET rate = ?', payload.rate)
    } },
  ],
})
  • An event reaches A only if A declares koffi as a satisfied dependency and subscribed to that (plugin, event). Emitting an event not in your emits is refused.
  • Like core event subscriptions, handlers run without a user — but unlike them they do receive the emitter's payload. Delivery is fire-and-forget on a short timeout; a slow subscriber never blocks the emitter.

Testing without a running TREK

createMockHost gives you a ctx that enforces the same permission model, so a test can prove your plugin degrades gracefully when a grant is missing:

import { createMockHost } from 'trek-plugin-sdk/testing'

const { ctx, broadcasts } = createMockHost({
  grants: ['db:read:trips'],
  trips: { 1: { members: [42], data: { id: 1, name: 'Japan' } } },
})
await ctx.trips.getById(1, 42)                        // ok — member
await expect(ctx.trips.getById(1, 99)).rejects…       // RESOURCE_FORBIDDEN
await expect(ctx.db.query('SELECT 1')).rejects…       // PERMISSION_DENIED (no db:own)

The mock db is a recorder — set queryResults for canned rows, or use an integration test for real SQL. To test inter-plugin calls, pass pluginExports: { koffi: { convert: (args) => … } } and assert on mock.emitted for anything your plugin publishes via ctx.events.emit.

Driving your plugin's handlers

createMockHost also gives you run(def) — the other half of a test. Where the ctx recorders (calls, broadcasts, emitted, notifications, scheduled) capture what your plugin read, run fires each entry point so you can assert what it did, all against the same mock ctx:

const host = createMockHost({ grants: ['jobs:run', 'hook:trip-card-provider'], actingUserId: 7 })
const app  = host.run(myPlugin)

await app.route({ method: 'GET', path: '/ping' })          // → the route's response
await app.job('refresh')                                    // fire a background job (userless)
await app.scheduled('daily', { tripId: 1 })                 // fire the `scheduled` handler
await app.event('place:created', { tripId: 1, entityId: 9 })// deliver a core event to `events`
await app.deleteUserData(42)                                // GDPR erasure handler
const dump  = await app.exportUserData(42)                  // GDPR export handler
const cards = await app.hook('tripCardProvider', 'getCards', [1, 2]) // any provider hook

host.scheduled.get('daily')                                 // timers the plugin armed via ctx.scheduler

A handler your plugin didn't declare throws a clear error (not a silent no-op), so a test catches a missing scheduled/deleteUserData/job before release.

Rules

  • No native modules (.node, binding.gyp, prebuilds/) — rejected at pack and install time.
  • Don't vendor trek-plugin-sdk — it's injected at runtime (devDependency only). Vendor any other runtime deps: TREK never runs npm install on a plugin.
  • Ship built JS in server/index.js and pre-built static files in client/. .ts and .map files are stripped by pack.
  • Declare every outbound host in egress[] whenever you use http:outbound.

Manifest reference (trek-plugin.json)

Field Type Notes
id string, required lowercase slug, ^[a-z][a-z0-9-]{2,39}$ (3–40 chars). Must match the directory name.
name string, required display name; also the page nav label.
version string, required semver (1.2.3, optional pre-release).
apiVersion number plugin API version (currently 1; PLUGIN_API_VERSION). Defaults to 1.
type string, required integration | page | widget | trip-page.
trek string supported TREK range, e.g. ">=3.2.0 <4.0.0". Its lower bound becomes minTrekVersion in the registry entry.
author string shown in the store.
description string one-line summary for the store.
icon string lucide-react icon name (default Blocks); used for the page nav entry.
homepage string project URL.
license string shown in the store detail (read from the manifest, not enforced).
nativeModules boolean must be false/absent — true is rejected.
permissions string[] see below.
egress string[] allowed outbound hosts; required (non-empty, no bare *) when any http:outbound permission is present.
capabilities.widget object { title, slot, defaultSize } — slot is sidebar (default), hero, place-detail, day-detail, or reservation-detail.
capabilities.tripPage object { replaces?, position? } for trip-page plugins — replaces names core planner tabs to hide while active (transports, buchungen, listen, finanzplan, dateien, collab; never plan), position is the tab's 0-based index in the bar (0–50; omitted = appended).
capabilities.provides string[] function names this plugin exposes to its dependents via ctx.plugins.call (see Talking to other plugins).
capabilities.emits string[] event names this plugin publishes to its dependents via ctx.events.emit.
requiredAddons string[] addon ids that must be enabled for the plugin to activate (see Dependencies).
pluginDependencies { id, version }[] other plugins (semver range) that must be installed + version-satisfied to activate.
settings array setting fields (below).

Permissions — the commonly-used core subset below; the full list of ~50 (all read/write scopes, the notify/ai/oauth brokers, every provider hook) lives in Plugin Permissions. Unknown values are rejected at activation.

Permission Grants
db:own ctx.db — your own SQLite file
db:read:trips ctx.trips.* (membership-checked, route handlers only)
db:read:packing ctx.packing.list(tripId) — a trip's packing items (membership-checked)
db:read:files ctx.files.list(tripId) — a trip's files, trash excluded (membership-checked)
db:read:costs ctx.costs.getByTrip / ctx.costs.listMine (Costs addon, route handlers only)
db:write:costs ctx.costs.create/update/delete (Costs addon + acting user's budget_edit)
db:write:places ctx.places.create/update/delete (acting user's place_edit)
db:write:days ctx.days.create/update/delete (acting user's day_edit)
db:write:itinerary ctx.itinerary.assign/unassign (acting user's day_edit)
db:write:trips ctx.trips.update (acting user's trip_edit)
db:write:packing ctx.packing.create/update/delete (acting user's packing_edit; private items stay owner-scoped)
db:meta ctx.meta.* — your own namespaced data on a trip/place/day/reservation/accommodation
db:read:users ctx.users.getById
events:subscribe receive core activity events via events: [...] (event name + tripId + a { entity, entityId } hint, plus a whitelisted entity snapshot when the plugin also holds the family's db:read:* grant; never a user)
hook:trip-card-provider hooks.tripCardProvider — small badges on the dashboard trip cards
jobs:run run declared background jobs on their cron schedule and ctx.scheduler runtime timers → scheduled handler (opt-in; no user, so trip reads are refused)
ws:broadcast:trip ctx.ws.broadcastToTrip
ws:broadcast:user ctx.ws.broadcastToUser
http:outbound or http:outbound:<host> outbound HTTP to egress[] hosts
hook:place-detail-provider hooks.placeDetailProvider — extra place rows TREK renders (see Provider hooks)
hook:trip-warning-provider hooks.warningProvider — validation warnings in the planner (see Provider hooks)
hook:table-contributor hooks.tableContributor — host-rendered columns/actions in the reservations, transports, places, day, costs, packing, files and todos views (see Provider hooks)
hook:map-marker-provider hooks.mapMarkerProvider — bounded markers on the trip map
hook:pdf-section-provider hooks.pdfSectionProvider — sections appended to the trip PDF export
hook:atlas-layer-provider hooks.atlasLayerProvider — per-user country tint layers on the Atlas map
hook:journal-entry-provider hooks.journalEntryProvider — extra rows on a journal entry card
hook:trip-card-provider hooks.tripCardProvider — small badges on the dashboard trip cards (getCards(tripIds, ctx) → { tripId, label, value?, icon?, tone?, url? }[]; host bounds every field + access-checks each tripId)
hook:user-data deleteUserData / exportUserData handlers — honour GDPR erasure (durable, retried) and data-export for a deleted/requesting user (userless; own db only)
hook:photo-provider / hook:calendar-source reserved (see Provider hooks)

There is no ws:broadcast:* — use ws:broadcast:trip and/or ws:broadcast:user explicitly.

Settings field (settings[]):

Key Notes
key required identifier; empty-key entries are dropped.
label form label.
input_type snake_case; e.g. text (default), password, number, select.
scope instance (default) or user.
required boolean.
secret boolean — encrypted at rest, decrypted only into ctx.config.
placeholder, hint form hints.
options [{ value, label }] for select inputs.
oauth { initPath, callbackPath } for OAuth flows.

Page nav: the host builds a page plugin's nav entry from the top-level name and icon. create-trek-plugin also scaffolds a capabilities.nav block, but the installed-manifest parser only consumes capabilities.widget — set name/icon to control the nav entry.

See Plugin Permissions for the full permission model.

The trek-plugin CLI

Run npx trek-plugin-sdk with no command in a terminal and you get an interactive menu (create / dev / validate / pack / publish, with signing and registry-entry commands under Advanced…); it just picks which command to run, then that command prompts for whatever it needs. Pass a command explicitly to skip the menu (and for scripts/CI).

Author commands (from trek-plugin-sdk):

# 1. Manifest + layout checks (a subset of the registry CI — CI additionally
#    verifies the GitHub release exists, the artifact sha256, and the README
#    over the network).
trek-plugin validate [dir]

# 2. Build plugin.zip in the installer's exact layout. Prints sha256 + byte size,
#    refuses native binaries, enforces the same size limits (25MB/file, 50MB total).
#    Ships trek-plugin.json, README.md, LICENSE(.md), package.json + server/ + client/.
#    docs/ is intentionally NOT shipped — the store fetches docs/screenshot.png
#    from your repo. --json prints a machine-readable result.
trek-plugin pack [dir] [--out plugin.zip] [--json]

# 3. Emit the ready-to-PR registry entry: commitSha (resolved from the git tag),
#    downloadUrl, sha256, size and minTrekVersion (derived from the manifest
#    'trek' range) all computed for you. --merge prepends a new version onto an
#    existing entry (the update case, kept newest-first).
trek-plugin entry --repo owner/name --tag vX.Y.Z [--zip plugin.zip] [--merge entry.json] [--out file]

# 4. One shot: pack -> create the GitHub release (via gh) -> print the entry.
trek-plugin release [dir] --repo owner/name --tag vX.Y.Z

To publish, open a PR that adds the emitted JSON as registry/plugins/<id>.json in the TREK-Plugins registry.

Registry & publishing

  • No reserved namespaces — any unique slug id is accepted. (A tiny set of ids like registry/install/rescan is blocked only because they'd collide with admin API routes.)
  • Owner-binding still prevents anyone but the original author from repointing an existing id to a different repo.
  • Optional author signing: an entry may carry authorPublicKey (stable, TOFU-pinned on first install) and each version a signature over the artifact bytes. Unsigned plugins install on sha256 alone; a plugin that was signed can't later go unsigned or swap its key without an explicit admin re-trust.

Full walkthrough: Publishing a Plugin. Overview: Plugins.